font
Filter
×
2025
Deshmukh, Jayati; Liang, Zijie; Yazdanpanah, Vahid; Stein, Sebastian; Ramchurn, Sarvalpali D.
Serious games for ethical preference elicitation Proceedings Article
In: AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25), 2025.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton498743,
title = {Serious games for ethical preference elicitation},
author = {Jayati Deshmukh and Zijie Liang and Vahid Yazdanpanah and Sebastian Stein and Sarvalpali D. Ramchurn},
url = {https://eprints.soton.ac.uk/498743/},
year = {2025},
date = {2025-05-01},
booktitle = {AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25)},
abstract = {Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.
2024
Thavanesan, Navamayooran; Farahi, Arya; Parfitt, Charlotte; Belkhatir, Zehor; Azim, Tayyaba; Vallejos, Elvira Perez; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Insights from explainable AI in oesophageal cancer team decisions Journal Article
In: Computers in Biology and Medicine, vol. 180, 2024, (For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.).
Abstract | Links | BibTeX | Tags: Decision-making, machine learning, Multidisciplinary teams, Oesophageal cancer
@article{soton493238,
title = {Insights from explainable AI in oesophageal cancer team decisions},
author = {Navamayooran Thavanesan and Arya Farahi and Charlotte Parfitt and Zehor Belkhatir and Tayyaba Azim and Elvira Perez Vallejos and Zoë Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/493238/},
year = {2024},
date = {2024-08-01},
journal = {Computers in Biology and Medicine},
volume = {180},
abstract = {ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>},
note = {For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.},
keywords = {Decision-making, machine learning, Multidisciplinary teams, Oesophageal cancer},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>
Naiseh, Mohammad; Webb, Catherine; Underwood, Tim; Ramchurn, Gopal; Walters, Zoe; Thavanesan, Navamayooran; Vigneswaran, Ganesh
XAI for group-AI interaction: towards collaborative and inclusive explanations Proceedings Article
In: Longo, Luca; Liu, Weiru; Montavon, Gregoire (Ed.): Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024), pp. 249–256, CEUR Workshop Proceedings, 2024.
Abstract | Links | BibTeX | Tags: Explainable AI, Group-AI Interaction, Interaction Design
@inproceedings{soton497829,
title = {XAI for group-AI interaction: towards collaborative and inclusive explanations},
author = {Mohammad Naiseh and Catherine Webb and Tim Underwood and Gopal Ramchurn and Zoe Walters and Navamayooran Thavanesan and Ganesh Vigneswaran},
editor = {Luca Longo and Weiru Liu and Gregoire Montavon},
url = {https://eprints.soton.ac.uk/497829/},
year = {2024},
date = {2024-07-01},
booktitle = {Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024)},
volume = {3793},
pages = {249–256},
publisher = {CEUR Workshop Proceedings},
abstract = {ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>},
keywords = {Explainable AI, Group-AI Interaction, Interaction Design},
pubstate = {published},
tppubtype = {inproceedings}
}
ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>
Early, Joseph Arthur
Interpretable multiple instance learning PhD Thesis
University of Southampton, 2024.
Abstract | Links | BibTeX | Tags:
@phdthesis{soton490767,
title = {Interpretable multiple instance learning},
author = {Joseph Arthur Early},
url = {https://eprints.soton.ac.uk/490767/},
year = {2024},
date = {2024-06-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>
Kiden, Sarah; Stahl, Bernd; Townsend, Beverley; Maple, Carsten; Vincent, Charles; Sampson, Fraser; Gilbert, Geoff; Smith, Helen; Deshmukh, Jayati; Ross, Jen; Williams, Jennifer; Rincon, Jesus Martinez; Lisinska, Justyna; O?Shea, Karen; Abreu, Márjory Da Costa; Bencomo, Nelly; Deb, Oishi; Winter, Peter; Li, Phoebe; Torr, Philip; Lau, Pin Lean; Iniesta, Raquel; Ramchurn, Gopal; Stein, Sebastian; Yazdanpanah, Vahid
Responsible AI governance: A response to UN interim report on governing AI for humanity Technical Report
no. 10.5258/SOTON/PP0057, 2024.
@techreport{soton488908,
title = {Responsible AI governance: A response to UN interim report on governing AI for humanity},
author = {Sarah Kiden and Bernd Stahl and Beverley Townsend and Carsten Maple and Charles Vincent and Fraser Sampson and Geoff Gilbert and Helen Smith and Jayati Deshmukh and Jen Ross and Jennifer Williams and Jesus Martinez Rincon and Justyna Lisinska and Karen O?Shea and Márjory Da Costa Abreu and Nelly Bencomo and Oishi Deb and Peter Winter and Phoebe Li and Philip Torr and Pin Lean Lau and Raquel Iniesta and Gopal Ramchurn and Sebastian Stein and Vahid Yazdanpanah},
url = {https://eprints.soton.ac.uk/488908/},
year = {2024},
date = {2024-03-01},
number = {10.5258/SOTON/PP0057},
publisher = {Public Policy, University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Abioye, Ayodeji O.; Hunt, William; Gu, Yue; Schneiders, Eike; Naiseh, Mohammad; Fischer, Joel E.; Ramchurn, Sarvapali D.; Soorati, Mohammad D.; Archibald, Blair; Sevegnani, Michele
The effect of predictive formal modelling at runtime on performance in human-swarm interaction Proceedings Article
In: HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction, pp. 172?176, Association for Computing Machinery, 2024, (Publisher Copyright: © 2024 Copyright held by the owner/author(s)).
Abstract | Links | BibTeX | Tags: Human-Robot Interaction (HRI), Human-Swarm Interaction (HSI), Predictive Formal Modelling (PFM), Task Performance
@inproceedings{soton488273,
title = {The effect of predictive formal modelling at runtime on performance in human-swarm interaction},
author = {Ayodeji O. Abioye and William Hunt and Yue Gu and Eike Schneiders and Mohammad Naiseh and Joel E. Fischer and Sarvapali D. Ramchurn and Mohammad D. Soorati and Blair Archibald and Michele Sevegnani},
url = {https://eprints.soton.ac.uk/488273/},
year = {2024},
date = {2024-03-01},
booktitle = {HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction},
pages = {172?176},
publisher = {Association for Computing Machinery},
abstract = {Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.},
note = {Publisher Copyright:
© 2024 Copyright held by the owner/author(s)},
keywords = {Human-Robot Interaction (HRI), Human-Swarm Interaction (HSI), Predictive Formal Modelling (PFM), Task Performance},
pubstate = {published},
tppubtype = {inproceedings}
}
Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.
Soorati, Mohammad D.; Naiseh, Mohammad; Hunt, William; Parnell, Katie; Clark, Jediah; Ramchurn, Sarvapali D.
Enabling trustworthiness in human-swarm systems through a digital twin Book Section
In: Dasgupta, Prithviraj; Llinas, James; Gillespie, Tony; Fouse, Scott; Lawless, William; Mittu, Ranjeev; Sofge, Donlad (Ed.): Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams, pp. 93–125, Academic Press, 2024, (Publisher Copyright: © 2024 Elsevier Inc. All rights reserved.).
Abstract | Links | BibTeX | Tags: Digital twin, Explainability, Human-swarm interaction, Trustworthy Autonomous Systems, User-centered design
@incollection{soton491769,
title = {Enabling trustworthiness in human-swarm systems through a digital twin},
author = {Mohammad D. Soorati and Mohammad Naiseh and William Hunt and Katie Parnell and Jediah Clark and Sarvapali D. Ramchurn},
editor = {Prithviraj Dasgupta and James Llinas and Tony Gillespie and Scott Fouse and William Lawless and Ranjeev Mittu and Donlad Sofge},
url = {https://eprints.soton.ac.uk/491769/},
year = {2024},
date = {2024-02-01},
booktitle = {Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams},
pages = {93–125},
publisher = {Academic Press},
abstract = {Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.},
note = {Publisher Copyright:
© 2024 Elsevier Inc. All rights reserved.},
keywords = {Digital twin, Explainability, Human-swarm interaction, Trustworthy Autonomous Systems, User-centered design},
pubstate = {published},
tppubtype = {incollection}
}
Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.
Thavanesan, Navamayooran; Parfitt, Charlotte; Bodala, Indu; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy; Vigneswaran, Ganesh
Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction Miscellaneous
2024.
Abstract | Links | BibTeX | Tags:
@misc{soton497828,
title = {Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction},
author = {Navamayooran Thavanesan and Charlotte Parfitt and Indu Bodala and Zoë Walters and Sarvapali Ramchurn and Timothy Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/497828/},
year = {2024},
date = {2024-01-01},
journal = {European Journal of Surgical Oncology},
volume = {50},
number = {1},
abstract = {Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035
Kelly, Thomas Graham; Soorati, Mohammad; Zauner, Klaus-Peter; Ramchurn, Gopal; Tarapore, Danesh
Trade-offs of dynamic control structure in human-swarm systems Proceedings Article
In: The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024, 2024.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton492838,
title = {Trade-offs of dynamic control structure in human-swarm systems},
author = {Thomas Graham Kelly and Mohammad Soorati and Klaus-Peter Zauner and Gopal Ramchurn and Danesh Tarapore},
url = {https://eprints.soton.ac.uk/492838/},
year = {2024},
date = {2024-01-01},
booktitle = {The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024},
abstract = {Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.
2023
Early, Joseph; Deweese, Ying-Jung Chen; Evers, Christine; Ramchurn, Sarvapali
Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation Journal Article
In: Environmental Data Science, vol. 2, pp. 18, 2023.
Abstract | Links | BibTeX | Tags:
@article{soton490766,
title = {Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation},
author = {Joseph Early and Ying-Jung Chen Deweese and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/490766/},
year = {2023},
date = {2023-12-01},
journal = {Environmental Data Science},
volume = {2},
pages = {18},
abstract = {Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.
Rigley, Eryn; Bentley, Caitlin; Krook, Joshua; Ramchurn, Gopal
Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries Journal Article
In: Global Policy, 2023, (Funding Information: This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.).
Abstract | Links | BibTeX | Tags:
@article{soton485727,
title = {Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries},
author = {Eryn Rigley and Caitlin Bentley and Joshua Krook and Gopal Ramchurn},
url = {https://eprints.soton.ac.uk/485727/},
year = {2023},
date = {2023-12-01},
journal = {Global Policy},
abstract = {ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>},
note = {Funding Information:
This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>
Singh, Lokesh; Ramchurn, Gopal
The effect of automated agents on individual performance under induced stress Proceedings Article
In: Kalra, Jay (Ed.): Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition), pp. 118–127, AHFE International, 2023.
Abstract | Links | BibTeX | Tags: Decision-making, Human-agent, Individual performance, Induced stress, Time pressure
@inproceedings{soton485655,
title = {The effect of automated agents on individual performance under induced stress},
author = {Lokesh Singh and Gopal Ramchurn},
editor = {Jay Kalra},
url = {https://eprints.soton.ac.uk/485655/},
year = {2023},
date = {2023-11-01},
booktitle = {Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition)},
pages = {118–127},
publisher = {AHFE International},
abstract = {Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.},
keywords = {Decision-making, Human-agent, Individual performance, Induced stress, Time pressure},
pubstate = {published},
tppubtype = {inproceedings}
}
Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, vol. 49, no. 11, 2023, (Publisher Copyright: © 2023 The Author(s)).
Abstract | Links | BibTeX | Tags: Artificial Intelligence, machine learning, Oesophageal cancer multidisciplinary team
@article{soton479497b,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-11-01},
journal = {European Journal of Surgical Oncology},
volume = {49},
number = {11},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
© 2023 The Author(s)},
keywords = {Artificial Intelligence, machine learning, Oesophageal cancer multidisciplinary team},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub Response Miscellaneous
2023.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, chatgpt, european law, generative ai, Large Language Models, law and technology, technology policy
@misc{soton481740,
title = {AI large language models inquiry: TASHub Response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {Artificial Intelligence, chatgpt, european law, generative ai, Large Language Models, law and technology, technology policy},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub response Miscellaneous
2023.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, chatgpt, european law, generative ai, Large Language Models, law and technology, technology policy
@misc{soton481740b,
title = {AI large language models inquiry: TASHub response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {Artificial Intelligence, chatgpt, european law, generative ai, Large Language Models, law and technology, technology policy},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, 2023, (Publisher Copyright: copyright 2023 The Author(s)).
Abstract | Links | BibTeX | Tags: Artificial Intelligence, machine learning, Oesophageal cancer multidisciplinary team
@article{soton479497,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-07-01},
journal = {European Journal of Surgical Oncology},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
copyright 2023 The Author(s)},
keywords = {Artificial Intelligence, machine learning, Oesophageal cancer multidisciplinary team},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
Abioye, Ayodeji
University of Southampton, 2023.
Abstract | Links | BibTeX | Tags:
@phdthesis{soton479472,
title = {Multimodal speech and visual gesture control interface technique for small unmanned multirotor aircraft},
author = {Ayodeji Abioye},
url = {https://eprints.soton.ac.uk/479472/},
year = {2023},
date = {2023-07-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>
Abioye, Ayodeji; Naiseh, Mohammad; Hunt, William; Clark, Jediah R; Ramchurn, Sarvapali D; Soorati, Mohammad
The effect of data visualisation quality and task density on human-swarm interaction Proceedings Article
In: Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), IEEE, 2023.
Abstract | Links | BibTeX | Tags: featured_publication
@inproceedings{soton479970,
title = {The effect of data visualisation quality and task density on human-swarm interaction},
author = {Ayodeji Abioye and Mohammad Naiseh and William Hunt and Jediah R Clark and Sarvapali D Ramchurn and Mohammad Soorati},
url = {https://eprints.soton.ac.uk/479970/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
booktitle = {Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
publisher = {IEEE},
abstract = {Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.},
keywords = {featured_publication},
pubstate = {published},
tppubtype = {inproceedings}
}
Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.
Krook, Joshua; McAuley, Derek; Anderson, Stuart; Downer, John; Winter, Peter; Ramchurn, Sarvapali D
AI Foundation Models: initial review, CMA Consultation, TAS Hub Response Miscellaneous
2023.
Links | BibTeX | Tags: Artificial Intelligence, Competition policy, featured_publication, Foundation Models, Large Language Models, markets
@misc{soton477553,
title = {AI Foundation Models: initial review, CMA Consultation, TAS Hub Response},
author = {Joshua Krook and Derek McAuley and Stuart Anderson and John Downer and Peter Winter and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477553/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
publisher = {University of Southampton},
keywords = {Artificial Intelligence, Competition policy, featured_publication, Foundation Models, Large Language Models, markets},
pubstate = {published},
tppubtype = {misc}
}
Krook, Joshua; Downer, John; Winter, Peter; Williams, Jennifer; Ives, Jonathan; Bratu, Roxana; Sheir, Stephanie; Williams, Robin; Anderson, Stuart; Li, Phoebe; Ramamoorthy, Subramanian; Ramchurn, Sarvapali
AI regulation: a pro-innovation approach ? policy proposals: TASHub Response Miscellaneous
2023.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Consultation, innovation, Regulation, Trustworthy Autonomous Systems
@misc{soton478329,
title = {AI regulation: a pro-innovation approach ? policy proposals: TASHub Response},
author = {Joshua Krook and John Downer and Peter Winter and Jennifer Williams and Jonathan Ives and Roxana Bratu and Stephanie Sheir and Robin Williams and Stuart Anderson and Phoebe Li and Subramanian Ramamoorthy and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478329/},
year = {2023},
date = {2023-06-01},
publisher = {University of Southampton},
abstract = {Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence},
keywords = {Artificial Intelligence, Consultation, innovation, Regulation, Trustworthy Autonomous Systems},
pubstate = {published},
tppubtype = {misc}
}
Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence
Hunt, William; Ryan, Jack; Abioye, Ayodeji O; Ramchurn, Sarvapali D; Soorati, Mohammad D
Demonstrating performance benefits of human-swarm teaming Proceedings Article
In: Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems, pp. 3062–3064, International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS), 2023.
Abstract | Links | BibTeX | Tags: featured_publication
@inproceedings{soton479903,
title = {Demonstrating performance benefits of human-swarm teaming},
author = {William Hunt and Jack Ryan and Ayodeji O Abioye and Sarvapali D Ramchurn and Mohammad D Soorati},
url = {https://eprints.soton.ac.uk/479903/},
year = {2023},
date = {2023-05-01},
urldate = {2023-05-01},
booktitle = {Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems},
pages = {3062–3064},
publisher = {International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS)},
abstract = {Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.},
keywords = {featured_publication},
pubstate = {published},
tppubtype = {inproceedings}
}
Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.
Worrawichaipat, Phuriwat; Gerding, Enrico; Kaparias, Ioannis; Ramchurn, Sarvapali
Multi-agent signal-less intersection management with dynamic platoon formation Proceedings Article
In: 22nd International Conference on Autonomous Agents and Multiagent Systems (29/05/23 - 02/06/23), pp. 1542–1550, 2023.
Links | BibTeX | Tags: featured_publication
@inproceedings{soton478647,
title = {Multi-agent signal-less intersection management with dynamic platoon formation},
author = {Phuriwat Worrawichaipat and Enrico Gerding and Ioannis Kaparias and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478647/},
year = {2023},
date = {2023-05-01},
urldate = {2023-05-01},
booktitle = {22nd International Conference on Autonomous Agents and Multiagent Systems (29/05/23 - 02/06/23)},
pages = {1542–1550},
keywords = {featured_publication},
pubstate = {published},
tppubtype = {inproceedings}
}
Everett, Gregory; Beal, Ryan J; Matthews, Tim; Early, Joseph; Norman, Timothy J; Ramchurn, Sarvapali D
Inferring player location in sports matches: multi-agent spatial imputation from limited observations Miscellaneous
2023.
Abstract | Links | BibTeX | Tags: cs.LG, cs.MA
@misc{soton477020,
title = {Inferring player location in sports matches: multi-agent spatial imputation from limited observations},
author = {Gregory Everett and Ryan J Beal and Tim Matthews and Joseph Early and Timothy J Norman and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477020/},
year = {2023},
date = {2023-02-01},
abstract = {Understanding agent behaviour in Multi-Agent Systems (MAS) is an important problem in domains such as autonomous driving, disaster response, and sports analytics. Existing MAS problems typically use uniform timesteps with observations for all agents. In this work, we analyse the problem of agent location imputation, specifically posed in environments with non-uniform timesteps and limited agent observability (textttchar12695% missing values). Our approach uses Long Short-Term Memory and Graph Neural Network components to learn temporal and inter-agent patterns to predict the location of all agents at every timestep. We apply this to the domain of football (soccer) by imputing the location of all players in a game from sparse event data (e.g., shots and passes). Our model estimates player locations to within textttchar1266.9m; a textttchar12662% reduction in error from the best performing baseline. This approach facilitates downstream analysis tasks such as player physical metrics, player coverage, and team pitch control. Existing solutions to these tasks often require optical tracking data, which is expensive to obtain and only available to elite clubs. By imputing player locations from easy to obtain event data, we increase the accessibility of downstream tasks.},
keywords = {cs.LG, cs.MA},
pubstate = {published},
tppubtype = {misc}
}
Understanding agent behaviour in Multi-Agent Systems (MAS) is an important problem in domains such as autonomous driving, disaster response, and sports analytics. Existing MAS problems typically use uniform timesteps with observations for all agents. In this work, we analyse the problem of agent location imputation, specifically posed in environments with non-uniform timesteps and limited agent observability (textttchar12695% missing values). Our approach uses Long Short-Term Memory and Graph Neural Network components to learn temporal and inter-agent patterns to predict the location of all agents at every timestep. We apply this to the domain of football (soccer) by imputing the location of all players in a game from sparse event data (e.g., shots and passes). Our model estimates player locations to within textttchar1266.9m; a textttchar12662% reduction in error from the best performing baseline. This approach facilitates downstream analysis tasks such as player physical metrics, player coverage, and team pitch control. Existing solutions to these tasks often require optical tracking data, which is expensive to obtain and only available to elite clubs. By imputing player locations from easy to obtain event data, we increase the accessibility of downstream tasks.
2022
Ahmed, Sarah; Azim, Tayyaba; Early, Joseph Arthur; Ramchurn, Sarvapali
Revisiting deep fisher vectors: using fisher information to improve object classification Proceedings Article
In: British Machine Vision Conference (21/11/22 - 24/11/22), 2022.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton471260,
title = {Revisiting deep fisher vectors: using fisher information to improve object classification},
author = {Sarah Ahmed and Tayyaba Azim and Joseph Arthur Early and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/471260/},
year = {2022},
date = {2022-11-01},
booktitle = {British Machine Vision Conference (21/11/22 - 24/11/22)},
abstract = {Although deep learning models have become the gold standard in achieving outstanding results on a large variety of computer vision and machine learning tasks, the use of kernel methods has still not gone out of trend because of its potential to beat deep learning performances at a number of occasions. Given the potential of kernel techniques, prior works have also proposed the use of hybrid approaches combining deep learning with kernel learning to complement their respective strengths and weaknesses. This work develops this idea further by introducing an improved version of Fisher kernels derived from the deep Boltzmann machines (DBM). Our improved deep Fisher kernel (IDFK) utilises an approximation of the Fisher information matrix to derive improved Fisher vectors. We show IDFK can be utilised to retain a high degree of class separability, making it appropriate for classification and retrieval tasks. The efficacy of the proposed approach is evaluated on three benchmark data sets: MNIST, USPS and Alphanumeric, showing an improvement in classification performance over existing kernel approaches, and comparable performance to deep learning methods, but with much reduced computational costs. Using explainable AI methods, we also demonstrate why our IDFK leads to better classification performance.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Although deep learning models have become the gold standard in achieving outstanding results on a large variety of computer vision and machine learning tasks, the use of kernel methods has still not gone out of trend because of its potential to beat deep learning performances at a number of occasions. Given the potential of kernel techniques, prior works have also proposed the use of hybrid approaches combining deep learning with kernel learning to complement their respective strengths and weaknesses. This work develops this idea further by introducing an improved version of Fisher kernels derived from the deep Boltzmann machines (DBM). Our improved deep Fisher kernel (IDFK) utilises an approximation of the Fisher information matrix to derive improved Fisher vectors. We show IDFK can be utilised to retain a high degree of class separability, making it appropriate for classification and retrieval tasks. The efficacy of the proposed approach is evaluated on three benchmark data sets: MNIST, USPS and Alphanumeric, showing an improvement in classification performance over existing kernel approaches, and comparable performance to deep learning methods, but with much reduced computational costs. Using explainable AI methods, we also demonstrate why our IDFK leads to better classification performance.
Yazdanpanah, Vahid; Gerding, Enrico; Stein, Sebastian; Dastani, Mehdi; Jonker, Catholijn M; Norman, Timothy; Ramchurn, Sarvapali
Reasoning About Responsibility in Autonomous Systems: Challenges and Opportunities Journal Article
In: AI & Society, 2022.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Citizen-Centric AI Systems, human-agent collectives, Human-Centred AI, Multiagent Responsibility Reasoning, Multiagent Systems, Trustworthy Autonomous Systems
@article{soton471971,
title = {Reasoning About Responsibility in Autonomous Systems: Challenges and Opportunities},
author = {Vahid Yazdanpanah and Enrico Gerding and Sebastian Stein and Mehdi Dastani and Catholijn M Jonker and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/471971/},
year = {2022},
date = {2022-11-01},
journal = {AI & Society},
abstract = {Ensuring the trustworthiness of autonomous systems and artificial intelligenceensuremath<br/ensuremath>is an important interdisciplinary endeavour. In this position paper, we argue thatensuremath<br/ensuremath>this endeavour will benefit from technical advancements in capturing various forms of responsibility, and we present a comprehensive research agenda to achieve this. In particular, we argue that ensuring the reliability of autonomous system can take advantage of technical approaches for quantifying degrees of responsibility and for coordinating tasks based on that. Moreover, we deem that, in certifying the legality of an AI system, formal and computationally implementable notions of responsibility, blame, accountability, and liability are applicable for addressing potential responsibility gaps (i.e., situations in which a group is responsible, but individuals? responsibility may be unclear). This is a call to enable AI systems themselves, as well as those involved in the design, monitoring, and governance of AI systems, to represent and reason about who can be seen as responsible in prospect (e.g., for completing a task in future) and who can be seen as responsible retrospectively (e.g., for a failure that has already occurred). To that end, in this work, we show that across all stages of the design, development, and deployment of Trustworthy Autonomous Systems (TAS), responsibility reasoning should play a key role. This position paper is the first step towards establishing a road-map and research agenda on how the notion of responsibility can provide novel solution concepts for ensuring the reliability and legality of TAS and, as a result, enables an effective embedding of AI technologies into society.},
keywords = {Artificial Intelligence, Citizen-Centric AI Systems, human-agent collectives, Human-Centred AI, Multiagent Responsibility Reasoning, Multiagent Systems, Trustworthy Autonomous Systems},
pubstate = {published},
tppubtype = {article}
}
Ensuring the trustworthiness of autonomous systems and artificial intelligenceensuremath<br/ensuremath>is an important interdisciplinary endeavour. In this position paper, we argue thatensuremath<br/ensuremath>this endeavour will benefit from technical advancements in capturing various forms of responsibility, and we present a comprehensive research agenda to achieve this. In particular, we argue that ensuring the reliability of autonomous system can take advantage of technical approaches for quantifying degrees of responsibility and for coordinating tasks based on that. Moreover, we deem that, in certifying the legality of an AI system, formal and computationally implementable notions of responsibility, blame, accountability, and liability are applicable for addressing potential responsibility gaps (i.e., situations in which a group is responsible, but individuals? responsibility may be unclear). This is a call to enable AI systems themselves, as well as those involved in the design, monitoring, and governance of AI systems, to represent and reason about who can be seen as responsible in prospect (e.g., for completing a task in future) and who can be seen as responsible retrospectively (e.g., for a failure that has already occurred). To that end, in this work, we show that across all stages of the design, development, and deployment of Trustworthy Autonomous Systems (TAS), responsibility reasoning should play a key role. This position paper is the first step towards establishing a road-map and research agenda on how the notion of responsibility can provide novel solution concepts for ensuring the reliability and legality of TAS and, as a result, enables an effective embedding of AI technologies into society.
Early, Joseph; Deweese, Ying-Jung; Evers, Christine; Ramchurn, Sarvapali
Scene-to-Patch earth observation: multiple instance learning for land cover classification Miscellaneous
2022, (14 pages total (4 main content; 2 acknowledgments + citations; 8 appendices); 8 figures (2 main; 6 appendix); published at "Tackling Climate Change with Machine Learning: Workshop at NeurIPS 2022").
Abstract | Links | BibTeX | Tags: cs.CV, cs.LG
@misc{soton472853,
title = {Scene-to-Patch earth observation: multiple instance learning for land cover classification},
author = {Joseph Early and Ying-Jung Deweese and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/472853/},
year = {2022},
date = {2022-11-01},
abstract = {Land cover classification (LCC), and monitoring how land use changes over time, is an important process in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation data for LCC rely on fully-annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of LCC. In this study, we propose Scene-to-Patch models: an alternative LCC approach utilising Multiple Instance Learning (MIL) that requires only high-level scene labels. This enables much faster development of new datasets whilst still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using LCC for different scenarios. On the DeepGlobe-LCC dataset, our approach outperforms non-MIL baselines on both scene- and patch-level prediction. This work provides the foundation for expanding the use of LCC in climate change mitigation methods for technology, government, and academia.},
note = {14 pages total (4 main content; 2 acknowledgments + citations; 8 appendices); 8 figures (2 main; 6 appendix); published at "Tackling Climate Change with Machine Learning: Workshop at NeurIPS 2022"},
keywords = {cs.CV, cs.LG},
pubstate = {published},
tppubtype = {misc}
}
Land cover classification (LCC), and monitoring how land use changes over time, is an important process in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation data for LCC rely on fully-annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of LCC. In this study, we propose Scene-to-Patch models: an alternative LCC approach utilising Multiple Instance Learning (MIL) that requires only high-level scene labels. This enables much faster development of new datasets whilst still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using LCC for different scenarios. On the DeepGlobe-LCC dataset, our approach outperforms non-MIL baselines on both scene- and patch-level prediction. This work provides the foundation for expanding the use of LCC in climate change mitigation methods for technology, government, and academia.
Parnell, Katie; Fischer, Joel E; Clark, Jediah R; Bodenmann, Adrian; Trigo, Maria Jose Galvez; Brito, Mario; Soorati, Mohammad Divband; Plant, Katherine; Ramchurn, Sarvapali
Trustworthy UAV relationships: Applying the Schema Action World taxonomy to UAVs and UAV swarm operations Journal Article
In: International Journal of Human-Computer Interaction, 2022.
Abstract | Links | BibTeX | Tags:
@article{soton468839,
title = {Trustworthy UAV relationships: Applying the Schema Action World taxonomy to UAVs and UAV swarm operations},
author = {Katie Parnell and Joel E Fischer and Jediah R Clark and Adrian Bodenmann and Maria Jose Galvez Trigo and Mario Brito and Mohammad Divband Soorati and Katherine Plant and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/468839/},
year = {2022},
date = {2022-07-01},
journal = {International Journal of Human-Computer Interaction},
abstract = {Human Factors play a significant role inthe development and integration of avionic systems to ensure that they are trusted and can be used effectively. As Unoccupied Aerial Vehicle (UAV) technology becomes increasingly important to the aviation domain this holds true. This study aims to gain an understanding of UAV operators?trust requirements when piloting UAVs by utilising a popular aviation interview methodology (Schema World Action Research Method), in combination with key questions on trust identified from the literature. Interviews were conducted with six UAVoperators, with a range of experience. This identified the importance of past experience to trust and the expectations that operators hold. Recommendations are made that target training to inform experience, in addition to the equipment, procedures and organisational standards that can aid in developing trustworthy systems. The methodology that was developed shows promise for capturing trust within human-automation interactions},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Human Factors play a significant role inthe development and integration of avionic systems to ensure that they are trusted and can be used effectively. As Unoccupied Aerial Vehicle (UAV) technology becomes increasingly important to the aviation domain this holds true. This study aims to gain an understanding of UAV operators?trust requirements when piloting UAVs by utilising a popular aviation interview methodology (Schema World Action Research Method), in combination with key questions on trust identified from the literature. Interviews were conducted with six UAVoperators, with a range of experience. This identified the importance of past experience to trust and the expectations that operators hold. Recommendations are made that target training to inform experience, in addition to the equipment, procedures and organisational standards that can aid in developing trustworthy systems. The methodology that was developed shows promise for capturing trust within human-automation interactions
Soorati, Mohammad Divband; Gerding, Enrico; Marchioni, Enrico; Naumov, Pavel; Norman, Timothy; Ramchurn, Sarvapali; Rastegari, Baharak; Sobey, Adam; Stein, Sebastian; Tarapore, Danesh; Yazdanpanah, Vahid; Zhang, Jie
From Intelligent Agents to Trustworthy Human-Centred Multiagent Systems Journal Article
In: AI Communications, 2022.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Distributed Artificial Intelligence, Intelligent Agents, Multiagent Systems, Trustworthy Autonomous Systems
@article{soton467975,
title = {From Intelligent Agents to Trustworthy Human-Centred Multiagent Systems},
author = {Mohammad Divband Soorati and Enrico Gerding and Enrico Marchioni and Pavel Naumov and Timothy Norman and Sarvapali Ramchurn and Baharak Rastegari and Adam Sobey and Sebastian Stein and Danesh Tarapore and Vahid Yazdanpanah and Jie Zhang},
url = {https://eprints.soton.ac.uk/467975/},
year = {2022},
date = {2022-07-01},
journal = {AI Communications},
abstract = {The Agents, Interaction and Complexity research group at the University of Southampton has a long track record of research in multiagent systems (MAS). We have made substantial scientific contributions across learning in MAS, game-theoretic techniques for coordinating agent systems, and formal methods for representation and reasoning. We highlight key results achieved by the group and elaborate on recent work and open research challenges in developing trustworthy autonomous systems and deploying human-centred AI systems that aim to support societal good.},
keywords = {Artificial Intelligence, Distributed Artificial Intelligence, Intelligent Agents, Multiagent Systems, Trustworthy Autonomous Systems},
pubstate = {published},
tppubtype = {article}
}
The Agents, Interaction and Complexity research group at the University of Southampton has a long track record of research in multiagent systems (MAS). We have made substantial scientific contributions across learning in MAS, game-theoretic techniques for coordinating agent systems, and formal methods for representation and reasoning. We highlight key results achieved by the group and elaborate on recent work and open research challenges in developing trustworthy autonomous systems and deploying human-centred AI systems that aim to support societal good.
Bossens, David; Ramchurn, Sarvapali; Tarapore, Danesh
Resilient robot teams: a review integrating decentralised control, change-detection, and learning Miscellaneous
2022.
Abstract | Links | BibTeX | Tags:
@misc{soton457101,
title = {Resilient robot teams: a review integrating decentralised control, change-detection, and learning},
author = {David Bossens and Sarvapali Ramchurn and Danesh Tarapore},
url = {https://eprints.soton.ac.uk/457101/},
year = {2022},
date = {2022-06-01},
journal = {Current Robotics Reports},
abstract = {Purpose of review: This paper reviews opportunities and challenges for decentralised control, change-detection, and learning in the context of resilient robot teams.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Recent findings: Exogenous fault detection methods can provide a generic detection or a specific diagnosis with a recovery solution. Robot teams can perform active and distributed sensing for detecting changes in the environment, including identifying and tracking dynamic anomalies, as well as collaboratively mapping dynamic environments. Resilient methods for decentralised control have been developed in learning perception-action-communication loops, multi-agent reinforcement learning, embodied evolution, offline evolution with online adaptation, explicit task allocation, and stigmergy in swarm robotics.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Summary: Remaining challenges for resilient robot teams are integrating change-detection and trial-and-error learning methods, obtaining reliable performance evaluations under constrained evaluation time, improving the safety of resilient robot teams, theoretical results demonstrating rapid adaptation to given environmental perturbations, and designing realistic and compelling case studies.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Purpose of review: This paper reviews opportunities and challenges for decentralised control, change-detection, and learning in the context of resilient robot teams.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Recent findings: Exogenous fault detection methods can provide a generic detection or a specific diagnosis with a recovery solution. Robot teams can perform active and distributed sensing for detecting changes in the environment, including identifying and tracking dynamic anomalies, as well as collaboratively mapping dynamic environments. Resilient methods for decentralised control have been developed in learning perception-action-communication loops, multi-agent reinforcement learning, embodied evolution, offline evolution with online adaptation, explicit task allocation, and stigmergy in swarm robotics.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Summary: Remaining challenges for resilient robot teams are integrating change-detection and trial-and-error learning methods, obtaining reliable performance evaluations under constrained evaluation time, improving the safety of resilient robot teams, theoretical results demonstrating rapid adaptation to given environmental perturbations, and designing realistic and compelling case studies.
Early, Joseph; Bewley, Tom; Evers, Christine; Ramchurn, Sarvapali
Non-markovian reward modelling from trajectory labels via interpretable multiple instance learning Journal Article
In: arXiv, 2022, (20 pages (9 main content; 2 references; 9 appendix). 11 figures (8 main content; 3 appendix)).
Abstract | Links | BibTeX | Tags: cs.AI, cs.LG
@article{soton458023,
title = {Non-markovian reward modelling from trajectory labels via interpretable multiple instance learning},
author = {Joseph Early and Tom Bewley and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/458023/},
year = {2022},
date = {2022-05-01},
journal = {arXiv},
abstract = {We generalise the problem of reward modelling (RM) for reinforcement learning (RL) to handle non-Markovian rewards. Existing work assumes that human evaluators observe each step in a trajectory independently when providing feedback on agent behaviour. In this work, we remove this assumption, extending RM to include hidden state information that captures temporal dependencies in human assessment of trajectories. We then show how RM can be approached as a multiple instance learning (MIL) problem, and develop new MIL models that are able to capture the time dependencies in labelled trajectories. We demonstrate on a range of RL tasks that our novel MIL models can reconstruct reward functions to a high level of accuracy, and that they provide interpretable learnt hidden information that can be used to train high-performing agent policies.},
note = {20 pages (9 main content; 2 references; 9 appendix). 11 figures (8 main content; 3 appendix)},
keywords = {cs.AI, cs.LG},
pubstate = {published},
tppubtype = {article}
}
We generalise the problem of reward modelling (RM) for reinforcement learning (RL) to handle non-Markovian rewards. Existing work assumes that human evaluators observe each step in a trajectory independently when providing feedback on agent behaviour. In this work, we remove this assumption, extending RM to include hidden state information that captures temporal dependencies in human assessment of trajectories. We then show how RM can be approached as a multiple instance learning (MIL) problem, and develop new MIL models that are able to capture the time dependencies in labelled trajectories. We demonstrate on a range of RL tasks that our novel MIL models can reconstruct reward functions to a high level of accuracy, and that they provide interpretable learnt hidden information that can be used to train high-performing agent policies.
Buermann, Jan; Georgiev, Dimitar; Gerding, Enrico; Hill, Lewis; Malik, Obaid; Pop, Alexandru; Pun, Matthew; Ramchurn, Sarvapali; Salisbury, Elliot; Stojanovic, Ivan
An agent-based simulator for maritime transport decarbonisation: Demonstration track Proceedings Article
In: 21st International Conference on Autonomous Agents and Multiagent Systems (09/05/22 - 13/05/22), pp. 1890–1892, 2022.
Abstract | Links | BibTeX | Tags: Agent-Based Modelling and Simulation: Applications & Analysis, Emergent Behaviour, Integration of Agent-Based and Other Technologies, Simulation Techniques, Tools and Platforms
@inproceedings{soton456716,
title = {An agent-based simulator for maritime transport decarbonisation: Demonstration track},
author = {Jan Buermann and Dimitar Georgiev and Enrico Gerding and Lewis Hill and Obaid Malik and Alexandru Pop and Matthew Pun and Sarvapali Ramchurn and Elliot Salisbury and Ivan Stojanovic},
url = {https://eprints.soton.ac.uk/456716/},
year = {2022},
date = {2022-05-01},
booktitle = {21st International Conference on Autonomous Agents and Multiagent Systems (09/05/22 - 13/05/22)},
pages = {1890–1892},
abstract = {Greenhouse gas (GHG) emission reduction is an important and necessary goal; currently, different policies to reduce GHG emissions in maritime transport are being discussed. Amongst policies, like carbon taxes or carbon intensity targets, it is hard to determine which policies can successfully reduce GHG emissions while allowing the industry to be profitable. We introduce an agent-based maritime transport simulator to investigate the effectiveness of two decarbonisation policies by simulating a maritime transport operator?s trade pattern and fleet make-up changes as a reaction to taxation and fixed targets. This first of its kind simulator allows to compare and quantify the difference of carbon reduction policies and how they affect shipping operations.},
keywords = {Agent-Based Modelling and Simulation: Applications & Analysis, Emergent Behaviour, Integration of Agent-Based and Other Technologies, Simulation Techniques, Tools and Platforms},
pubstate = {published},
tppubtype = {inproceedings}
}
Greenhouse gas (GHG) emission reduction is an important and necessary goal; currently, different policies to reduce GHG emissions in maritime transport are being discussed. Amongst policies, like carbon taxes or carbon intensity targets, it is hard to determine which policies can successfully reduce GHG emissions while allowing the industry to be profitable. We introduce an agent-based maritime transport simulator to investigate the effectiveness of two decarbonisation policies by simulating a maritime transport operator?s trade pattern and fleet make-up changes as a reaction to taxation and fixed targets. This first of its kind simulator allows to compare and quantify the difference of carbon reduction policies and how they affect shipping operations.
Rigas, Emmanouil S; Gerding, Enrico H; Stein, Sebastian; Ramchurn, Sarvapali D; Bassiliades, Nick
Mechanism design for efficient offline and online allocation of electric vehicles to charging stations Journal Article
In: Energies, vol. 15, no. 5, 2022, (Funding Information: Funding: This research study was co-financed by Greece and the European Union (European Social Fund–ESF) through the Operational Programme ?Human Resources Development, Education and Lifelong Learning? in the context of the project ?Reinforcement of Postdoctoral Researchers-2nd Cycle? (MIS-5033021), implemented by State Scholarships Foundation (IKY). Copyright 2022 Elsevier B.V., All rights reserved.).
Abstract | Links | BibTeX | Tags: Charging, Electric Vehicles, Fixed price, mechanism design, Scheduling, VCG
@article{soton455806,
title = {Mechanism design for efficient offline and online allocation of electric vehicles to charging stations},
author = {Emmanouil S Rigas and Enrico H Gerding and Sebastian Stein and Sarvapali D Ramchurn and Nick Bassiliades},
url = {https://eprints.soton.ac.uk/455806/},
year = {2022},
date = {2022-03-01},
journal = {Energies},
volume = {15},
number = {5},
abstract = {ensuremath<pensuremath>The industry related to electric vehicles (EVs) has seen a substantial increase in recent years, as such vehicles have the ability to significantly reduce total COensuremath<subensuremath>2ensuremath</subensuremath> emissions and the related global warming effect. In this paper, we focus on the problem of allocating EVs to charging stations, scheduling and pricing their charging. Specifically, we developed a Mixed Integer Program (MIP) which executes offline and optimally allocates EVs to charging stations. On top, we propose two alternative mechanisms to price the electricity the EVs charge. The first mechanism is a typical fixed-price one, while the second is a variation of the Vickrey?Clark?Groves (VCG) mechanism. We also developed online solutions that incrementally call the MIP-based algorithm and solve it for branches of EVs. In all cases, the EVs? aim is to minimize the price to pay and the impact on their driving schedule, acting as self-interested agents. We conducted a thorough empirical evaluation of our mechanisms and we observed that they had satisfactory scalability. Additionally, the VCG mechanism achieved an up to 2.2% improvement in terms of the number of vehicles that were charged compared to the fixed-price one and, in cases where the stations were congested, it calculated higher prices for the EVs and provided a higher profit for the stations, but lower utility to the EVs. However, in a theoretical evaluation, we proved that the variant of the VCG mechanism being proposed in this paper still guaranteed truthful reporting of the EVs? preferences. In contrast, the fixed-price one was found to be vulnerable to agents? strategic behavior as non-truthful EVs can charge instead of truthful ones. Finally, we observed the online algorithms to be, on average, at 95.6% of the offline ones in terms of the average number of serviced EVs.ensuremath</pensuremath>},
note = {Funding Information:
Funding: This research study was co-financed by Greece and the European Union (European Social Fund–ESF) through the Operational Programme ?Human Resources Development, Education and Lifelong Learning? in the context of the project ?Reinforcement of Postdoctoral Researchers-2nd Cycle? (MIS-5033021), implemented by State Scholarships Foundation (IKY).
Copyright 2022 Elsevier B.V., All rights reserved.},
keywords = {Charging, Electric Vehicles, Fixed price, mechanism design, Scheduling, VCG},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>The industry related to electric vehicles (EVs) has seen a substantial increase in recent years, as such vehicles have the ability to significantly reduce total COensuremath<subensuremath>2ensuremath</subensuremath> emissions and the related global warming effect. In this paper, we focus on the problem of allocating EVs to charging stations, scheduling and pricing their charging. Specifically, we developed a Mixed Integer Program (MIP) which executes offline and optimally allocates EVs to charging stations. On top, we propose two alternative mechanisms to price the electricity the EVs charge. The first mechanism is a typical fixed-price one, while the second is a variation of the Vickrey?Clark?Groves (VCG) mechanism. We also developed online solutions that incrementally call the MIP-based algorithm and solve it for branches of EVs. In all cases, the EVs? aim is to minimize the price to pay and the impact on their driving schedule, acting as self-interested agents. We conducted a thorough empirical evaluation of our mechanisms and we observed that they had satisfactory scalability. Additionally, the VCG mechanism achieved an up to 2.2% improvement in terms of the number of vehicles that were charged compared to the fixed-price one and, in cases where the stations were congested, it calculated higher prices for the EVs and provided a higher profit for the stations, but lower utility to the EVs. However, in a theoretical evaluation, we proved that the variant of the VCG mechanism being proposed in this paper still guaranteed truthful reporting of the EVs? preferences. In contrast, the fixed-price one was found to be vulnerable to agents? strategic behavior as non-truthful EVs can charge instead of truthful ones. Finally, we observed the online algorithms to be, on average, at 95.6% of the offline ones in terms of the average number of serviced EVs.ensuremath</pensuremath>
Early, Joseph; Evers, Christine; Ramchurn, Sarvapali
Model agnostic interpretability for multiple instance learning Proceedings Article
In: International Conference on Learning Representations 2022 (25/04/22 - 29/04/22), 2022, (25 pages (9 content, 2 acknowledgement + references, 14 appendix). 16 figures (3 main content, 13 appendix). Submitted and accepted to ICLR 22, see http://openreview.net/forum?id=KSSfF5lMIAg . Revision: added additional acknowledgements).
Abstract | Links | BibTeX | Tags: cs.AI, cs.LG
@inproceedings{soton454952,
title = {Model agnostic interpretability for multiple instance learning},
author = {Joseph Early and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/454952/},
year = {2022},
date = {2022-01-01},
booktitle = {International Conference on Learning Representations 2022 (25/04/22 - 29/04/22)},
abstract = {In Multiple Instance Learning (MIL), models are trained using bags of instances, where only a single label is provided for each bag. A bag label is often only determined by a handful of key instances within a bag, making it difficult to interpret what information a classifier is using to make decisions. In this work, we establish the key requirements for interpreting MIL models. We then go on to develop several model-agnostic approaches that meet these requirements. Our methods are compared against existing inherently interpretable MIL models on several datasets, and achieve an increase in interpretability accuracy of up to 30%. We also examine the ability of the methods to identify interactions between instances and scale to larger datasets, improving their applicability to real-world problems.},
note = {25 pages (9 content, 2 acknowledgement + references, 14 appendix). 16 figures (3 main content, 13 appendix). Submitted and accepted to ICLR 22, see http://openreview.net/forum?id=KSSfF5lMIAg . Revision: added additional acknowledgements},
keywords = {cs.AI, cs.LG},
pubstate = {published},
tppubtype = {inproceedings}
}
In Multiple Instance Learning (MIL), models are trained using bags of instances, where only a single label is provided for each bag. A bag label is often only determined by a handful of key instances within a bag, making it difficult to interpret what information a classifier is using to make decisions. In this work, we establish the key requirements for interpreting MIL models. We then go on to develop several model-agnostic approaches that meet these requirements. Our methods are compared against existing inherently interpretable MIL models on several datasets, and achieve an increase in interpretability accuracy of up to 30%. We also examine the ability of the methods to identify interactions between instances and scale to larger datasets, improving their applicability to real-world problems.
2021
Ramchurn, Sarvapali; Middleton, Stuart; McAuley, Derek; Webb, Helena; Hyde, Richard; Lisinska, Justyna
A Response to Draft Online Safety Bill: A call for evidence from the Joint Committee Technical Report
no. 10.18742/pub01-060, 2021.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Online Harms
@techreport{soton451428,
title = {A Response to Draft Online Safety Bill: A call for evidence from the Joint Committee},
author = {Sarvapali Ramchurn and Stuart Middleton and Derek McAuley and Helena Webb and Richard Hyde and Justyna Lisinska},
url = {https://eprints.soton.ac.uk/451428/},
year = {2021},
date = {2021-09-01},
number = {10.18742/pub01-060},
abstract = {This report is the Trustworthy Autonomous Hub (TAS-hub) response to the call for evidence from the Joint Committee on the Draft Online Safety Bill. The Joint Committee was established to consider the Government's draft Bill to establish a new regulatory framework to tackle harmful content online.},
keywords = {Artificial Intelligence, Online Harms},
pubstate = {published},
tppubtype = {techreport}
}
This report is the Trustworthy Autonomous Hub (TAS-hub) response to the call for evidence from the Joint Committee on the Draft Online Safety Bill. The Joint Committee was established to consider the Government's draft Bill to establish a new regulatory framework to tackle harmful content online.
Ramchurn, Sarvapali; Stein, Sebastian; Jennings, Nicholas R
Trustworthy human-AI partnerships Journal Article
In: iScience, vol. 24, no. 8, 2021.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Human-Computer Interaction, Sociology
@article{soton450597,
title = {Trustworthy human-AI partnerships},
author = {Sarvapali Ramchurn and Sebastian Stein and Nicholas R Jennings},
url = {https://eprints.soton.ac.uk/450597/},
year = {2021},
date = {2021-08-01},
journal = {iScience},
volume = {24},
number = {8},
abstract = {In this paper, we foreground some of the key research challenges that arise in the design of trustworthy human-AI partnerships. In particular, we focus on the challenges in designing human-AI partnerships that need to be addressed to help humans and organisations trust their machine counterparts individually or as a collective (e.g., as robot teams or groups of software agents). We also aim to identify the risks associated with human-AI partnerships and therefore determine the associated measures to mitigate these risks. By so doing, we will trigger new avenues of research that will address the key barriers to the adoption of AI-based systems more widely in our daily lives and in industry.},
keywords = {Artificial Intelligence, Human-Computer Interaction, Sociology},
pubstate = {published},
tppubtype = {article}
}
In this paper, we foreground some of the key research challenges that arise in the design of trustworthy human-AI partnerships. In particular, we focus on the challenges in designing human-AI partnerships that need to be addressed to help humans and organisations trust their machine counterparts individually or as a collective (e.g., as robot teams or groups of software agents). We also aim to identify the risks associated with human-AI partnerships and therefore determine the associated measures to mitigate these risks. By so doing, we will trigger new avenues of research that will address the key barriers to the adoption of AI-based systems more widely in our daily lives and in industry.
Ramchurn, Sarvapali; Mousavi, Mohammad Reza; Toliyat, Seyed Mohammad Hossein; Kleinman, Mark; Lisinska, Justyna; Sempreboni, Diego; Stein, Sebastian; Gerding, Enrico; Gomer, Richard; DÁmore, Francesco
The future of connected and automated mobility in the UK: call for evidence Technical Report
no. 10.5258/SOTON/P0097, 2021, (The UKRI TAS Hub assembles a team from the Universities of Southampton, Nottingham and King?s College London. The Hub sits at the centre of the pounds33M Trustworthy Autonomous Systems Programme, funded by the UKRI Strategic Priorities Fund. The role of the TAS Hub is to coordinate and work with six research nodes to establish a collaborative platform for the UK to enable the development of socially beneficial autonomous systems that are both trustworthy in principle and trusted in practice by individuals, society and government. Read more about the TAS Hub at https://www.tas.ac.uk/aboutus/overview/).
Abstract | Links | BibTeX | Tags:
@techreport{soton450228,
title = {The future of connected and automated mobility in the UK: call for evidence},
author = {Sarvapali Ramchurn and Mohammad Reza Mousavi and Seyed Mohammad Hossein Toliyat and Mark Kleinman and Justyna Lisinska and Diego Sempreboni and Sebastian Stein and Enrico Gerding and Richard Gomer and Francesco DÁmore},
editor = {Wassim Dbouk},
url = {https://eprints.soton.ac.uk/450228/},
year = {2021},
date = {2021-07-01},
number = {10.5258/SOTON/P0097},
publisher = {University of Southampton},
abstract = {This report is a response to the call for evidence from the Department for Business, Energy & Industrial Strategy and the Centre for Connected and Autonomous Vehicles on the future of connected and automated mobility in the UK.ensuremath<br/ensuremath>Executive Summary:Despite relative weaknesses in global collaboration and co-creation platforms, smart road and communication infrastructure, urban planning, and public awareness, the United Kingdom (UK) has a substantial strength in the area of Connected and Automated Mobility (CAM) by investing in research and innovation platforms for developing the underlying technologies, creating impact, and co-creation leading to innovative solutions. Many UK legal and policymaking initiatives in this domain are world leading. To sustain the UK?s leading position, we make the following recommendations:? The development of financial and policy-related incentive schemes for research and innovation in the foundations and applications of autonomous systems as well as schemes for proof of concepts, and commercialisation.? Supporting policy and standardisation initiatives as well as engagement and community-building activities to increase public awareness and trust.? Giving greater attention to integrating CAM/Connected Autonomous Shared Electric vehicles (CASE) policy with related government priorities for mobility, including supporting active transport and public transport, and improving air quality.? Further investment in updating liability and risk models and coming up with innovative liability schemes covering the Autonomous Vehicles (AVs) ecosystem.? Investing in training and retraining of the work force in the automotive, mobility, and transport sectors, particularly with skills concerningArtificial Intelligence (AI), software and computer systems, in order to ensure employability and an adequate response to the drastically changing industrial landscape},
note = {The UKRI TAS Hub assembles a team from the Universities of Southampton, Nottingham and King?s College London. The Hub sits at the centre of the pounds33M Trustworthy Autonomous Systems Programme, funded by the UKRI Strategic Priorities Fund.
The role of the TAS Hub is to coordinate and work with six research nodes to establish a collaborative platform for the UK to enable the development of socially beneficial autonomous systems that are both trustworthy in principle and trusted in practice by individuals, society and government. Read more about the TAS Hub at https://www.tas.ac.uk/aboutus/overview/},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
This report is a response to the call for evidence from the Department for Business, Energy & Industrial Strategy and the Centre for Connected and Autonomous Vehicles on the future of connected and automated mobility in the UK.ensuremath<br/ensuremath>Executive Summary:Despite relative weaknesses in global collaboration and co-creation platforms, smart road and communication infrastructure, urban planning, and public awareness, the United Kingdom (UK) has a substantial strength in the area of Connected and Automated Mobility (CAM) by investing in research and innovation platforms for developing the underlying technologies, creating impact, and co-creation leading to innovative solutions. Many UK legal and policymaking initiatives in this domain are world leading. To sustain the UK?s leading position, we make the following recommendations:? The development of financial and policy-related incentive schemes for research and innovation in the foundations and applications of autonomous systems as well as schemes for proof of concepts, and commercialisation.? Supporting policy and standardisation initiatives as well as engagement and community-building activities to increase public awareness and trust.? Giving greater attention to integrating CAM/Connected Autonomous Shared Electric vehicles (CASE) policy with related government priorities for mobility, including supporting active transport and public transport, and improving air quality.? Further investment in updating liability and risk models and coming up with innovative liability schemes covering the Autonomous Vehicles (AVs) ecosystem.? Investing in training and retraining of the work force in the automotive, mobility, and transport sectors, particularly with skills concerningArtificial Intelligence (AI), software and computer systems, in order to ensure employability and an adequate response to the drastically changing industrial landscape
Worrawichaipat, Phuriwat; Gerding, Enrico; Kaparias, Ioannis; Ramchurn, Sarvapali
Resilient intersection management with multi-vehicle collision avoidance Journal Article
In: Frontiers in Sustainable Cities, vol. 3, 2021.
Abstract | Links | BibTeX | Tags: Computer science, intersection management, multi-agent systems, simulation experiments, Transportation
@article{soton449675,
title = {Resilient intersection management with multi-vehicle collision avoidance},
author = {Phuriwat Worrawichaipat and Enrico Gerding and Ioannis Kaparias and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449675/},
year = {2021},
date = {2021-06-01},
journal = {Frontiers in Sustainable Cities},
volume = {3},
abstract = {In this paper, we propose a novel decentralised agent-based mechanism for road intersection management for connected autonomous vehicles. In our work we focus on road obstructions causing major traffic delays. In doing so, we propose the first decentralised mechanism able to maximise the overall vehicle throughput at intersections in the presence of obstructions. The distributed algorithm transfers most of the computational cost from the intersection manager to the driving agents, thereby improving scalability. Our realistic empirical experiments using SUMO show that, when an obstacle is located at the entrance or in the middle of the intersection, existing state of the art algorithms and traffic lights show a reduced throughput of 65?90% from the optimal point without obstructions while our mechanism can maintain the throughput upensuremath<br/ensuremath>Q7 to 94?99%.},
keywords = {Computer science, intersection management, multi-agent systems, simulation experiments, Transportation},
pubstate = {published},
tppubtype = {article}
}
In this paper, we propose a novel decentralised agent-based mechanism for road intersection management for connected autonomous vehicles. In our work we focus on road obstructions causing major traffic delays. In doing so, we propose the first decentralised mechanism able to maximise the overall vehicle throughput at intersections in the presence of obstructions. The distributed algorithm transfers most of the computational cost from the intersection manager to the driving agents, thereby improving scalability. Our realistic empirical experiments using SUMO show that, when an obstacle is located at the entrance or in the middle of the intersection, existing state of the art algorithms and traffic lights show a reduced throughput of 65?90% from the optimal point without obstructions while our mechanism can maintain the throughput upensuremath<br/ensuremath>Q7 to 94?99%.
Capezzuto, Luca; Tarapore, Danesh; Ramchurn, Sarvapali
Large-scale, dynamic and distributed coalition formation with spatial and†temporal constraints Journal Article
In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), pp. 108–125, 2021.
Abstract | Links | BibTeX | Tags:
@article{soton452050,
title = {Large-scale, dynamic and distributed coalition formation with spatial and†temporal constraints},
author = {Luca Capezzuto and Danesh Tarapore and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/452050/},
year = {2021},
date = {2021-05-01},
journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)},
pages = {108–125},
abstract = {The†Coalition Formation with Spatial and Temporal constraints Problem†(CFSTP) is a multi-agent task allocation problem in which few agents have to perform many tasks, each with its deadline and workload. To maximize the number of completed tasks, the agents need to cooperate by forming, disbanding and reforming coalitions. The original mathematical programming formulation of the CFSTP is difficult to implement, since it is lengthy and based on the problematic Big-M method. In this paper, we propose a compact and easy-to-implement formulation. Moreover, we design D-CTS, a distributed version of the state-of-the-art CFSTP algorithm. Using public London Fire Brigade records, we create a dataset with 347588 tasks and a test framework that simulates the mobilization of firefighters in dynamic environments. In problems with up†to 150 agents and 3000 tasks, compared to DSA-SDP, a state-of-the-art distributed algorithm, D-CTS completes†3.79%$pm$[42.22%,1.96%]†more tasks, and is one order of magnitude more efficient in terms of communication overhead and time complexity. D-CTS sets the first large-scale, dynamic and distributed CFSTP benchmark.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The†Coalition Formation with Spatial and Temporal constraints Problem†(CFSTP) is a multi-agent task allocation problem in which few agents have to perform many tasks, each with its deadline and workload. To maximize the number of completed tasks, the agents need to cooperate by forming, disbanding and reforming coalitions. The original mathematical programming formulation of the CFSTP is difficult to implement, since it is lengthy and based on the problematic Big-M method. In this paper, we propose a compact and easy-to-implement formulation. Moreover, we design D-CTS, a distributed version of the state-of-the-art CFSTP algorithm. Using public London Fire Brigade records, we create a dataset with 347588 tasks and a test framework that simulates the mobilization of firefighters in dynamic environments. In problems with up†to 150 agents and 3000 tasks, compared to DSA-SDP, a state-of-the-art distributed algorithm, D-CTS completes†3.79%$pm$[42.22%,1.96%]†more tasks, and is one order of magnitude more efficient in terms of communication overhead and time complexity. D-CTS sets the first large-scale, dynamic and distributed CFSTP benchmark.ensuremath<br/ensuremath>
Capezzuto, Luca; Tarapore, Danesh; Ramchurn, Sarvapali D
Anytime and efficient multi-agent coordination for disaster response Journal Article
In: SN Computer Science, vol. 2, no. 3, 2021.
Abstract | Links | BibTeX | Tags:
@article{soton467373,
title = {Anytime and efficient multi-agent coordination for disaster response},
author = {Luca Capezzuto and Danesh Tarapore and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/467373/},
year = {2021},
date = {2021-03-01},
journal = {SN Computer Science},
volume = {2},
number = {3},
abstract = {The Coalition Formation with Spatial and Temporal constraints Problem (CFSTP) is a multi-agent task allocation problem where the tasks are spatially distributed, with deadlines and workloads, and the number of agents is typically much smaller than the number of tasks. To maximise the number of completed tasks, the agents may have to schedule coalitions. The state-of-the-art CFSTP solver, the Coalition Formation with Look-Ahead (CFLA) algorithm, has two main limitations. First, its time complexity is exponential with the number of agents. Second, as we show, its look-ahead technique is not effective in real-world scenarios, such as open multi-agent systems, where new tasks can appear at any time. In this work, we study its design and define a variant, called Coalition Formation with Improved Look-Ahead (CFLA2), which achieves better performance. Since we cannot eliminate the limitations of CFLA in CFLA2, we also develop a novel algorithm to solve the CFSTP, the first to be simultaneously anytime, efficient and with convergence guarantee, called Cluster-based Task Scheduling (CTS). In tests where the look-ahead technique is highly effective, CTS completes up to 30% (resp. 10%) more tasks than CFLA (resp. CFLA2) while being up to 4 orders of magnitude faster. We also propose S-CTS, a simplified but parallel variant of CTS with even lower time complexity. Using scenarios generated by the RoboCup Rescue Simulation, we show that S-CTS is at most 10% less performing than high-performance algorithms such as Binary Max-Sum and DSA, but up to 2 orders of magnitude faster. Our results affirm CTS as the new state-of-the-art algorithm to solve the CFSTP.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The Coalition Formation with Spatial and Temporal constraints Problem (CFSTP) is a multi-agent task allocation problem where the tasks are spatially distributed, with deadlines and workloads, and the number of agents is typically much smaller than the number of tasks. To maximise the number of completed tasks, the agents may have to schedule coalitions. The state-of-the-art CFSTP solver, the Coalition Formation with Look-Ahead (CFLA) algorithm, has two main limitations. First, its time complexity is exponential with the number of agents. Second, as we show, its look-ahead technique is not effective in real-world scenarios, such as open multi-agent systems, where new tasks can appear at any time. In this work, we study its design and define a variant, called Coalition Formation with Improved Look-Ahead (CFLA2), which achieves better performance. Since we cannot eliminate the limitations of CFLA in CFLA2, we also develop a novel algorithm to solve the CFSTP, the first to be simultaneously anytime, efficient and with convergence guarantee, called Cluster-based Task Scheduling (CTS). In tests where the look-ahead technique is highly effective, CTS completes up to 30% (resp. 10%) more tasks than CFLA (resp. CFLA2) while being up to 4 orders of magnitude faster. We also propose S-CTS, a simplified but parallel variant of CTS with even lower time complexity. Using scenarios generated by the RoboCup Rescue Simulation, we show that S-CTS is at most 10% less performing than high-performance algorithms such as Binary Max-Sum and DSA, but up to 2 orders of magnitude faster. Our results affirm CTS as the new state-of-the-art algorithm to solve the CFSTP.
Ortega, Andre P; Ramchurn, Sarvapali; Tran-Thanh, Long; Merrett, Geoff
Partner selection in self-organised wireless sensor networks for opportunistic energy negotiation: A multi-armed bandit based approach Journal Article
In: Ad Hoc Networks, vol. 112, 2021.
Abstract | Links | BibTeX | Tags: Agent-based sensor network, Automated negotiation, Energy management, Multi-armed bandit based learning, Reinforcement Learning, Wireless sensor networks
@article{soton445733,
title = {Partner selection in self-organised wireless sensor networks for opportunistic energy negotiation: A multi-armed bandit based approach},
author = {Andre P Ortega and Sarvapali Ramchurn and Long Tran-Thanh and Geoff Merrett},
url = {https://eprints.soton.ac.uk/445733/},
year = {2021},
date = {2021-03-01},
journal = {Ad Hoc Networks},
volume = {112},
abstract = {The proliferation of ?Things? over a network creates the Internet of Things (IoT), where sensors integrate to collect data from the environment over long periods of time. The growth of IoT applications will inevitably involve co-locating multiple wireless sensor networks, each serving different applications with, possibly, different needs and constraints. Since energy is scarce in sensor nodes equipped with non-rechargeable batteries, energy harvesting technologies have been the focus of research in recent years. However, new problems arise as a result of their wide spatio-temporal variation. Such a shortcoming can be avoided if co-located networks cooperate with each other and share their available energy. Due to their unique characteristics and different owners, recently, we proposed a negotiation approach to deal with conflict of preferences. Unfortunately, negotiation can be impractical with a large number of participants, especially in an open environment. Given this, we introduce a new partner selection technique based on multi-armed bandits (MAB), that enables each node to learn the strategy that optimises its energy resources in the long term. Our results show that the proposed solution allows networks to repeatedly learn the current best energy partner in a dynamic environment. The performance of such a technique is evaluated through simulation and shows that a network can achieve an efficiency of 72% against the optimal strategy in the most challenging scenario studied in this work.},
keywords = {Agent-based sensor network, Automated negotiation, Energy management, Multi-armed bandit based learning, Reinforcement Learning, Wireless sensor networks},
pubstate = {published},
tppubtype = {article}
}
The proliferation of ?Things? over a network creates the Internet of Things (IoT), where sensors integrate to collect data from the environment over long periods of time. The growth of IoT applications will inevitably involve co-locating multiple wireless sensor networks, each serving different applications with, possibly, different needs and constraints. Since energy is scarce in sensor nodes equipped with non-rechargeable batteries, energy harvesting technologies have been the focus of research in recent years. However, new problems arise as a result of their wide spatio-temporal variation. Such a shortcoming can be avoided if co-located networks cooperate with each other and share their available energy. Due to their unique characteristics and different owners, recently, we proposed a negotiation approach to deal with conflict of preferences. Unfortunately, negotiation can be impractical with a large number of participants, especially in an open environment. Given this, we introduce a new partner selection technique based on multi-armed bandits (MAB), that enables each node to learn the strategy that optimises its energy resources in the long term. Our results show that the proposed solution allows networks to repeatedly learn the current best energy partner in a dynamic environment. The performance of such a technique is evaluated through simulation and shows that a network can achieve an efficiency of 72% against the optimal strategy in the most challenging scenario studied in this work.
Yazdanpanah, Vahid; Gerding, Enrico H; Stein, Sebastian; Dastani, Mehdi; Jonker, Catholijn M; Norman, Timothy J; Ramchurn, Sarvapali D
Responsibility ascription in trustworthy autonomous systems Proceedings Article
In: Embedding AI in Society (18/02/21 - 19/02/21), 2021.
Abstract | Links | BibTeX | Tags: Multiagent Systems, Reliable AI, Responsibility Reasoning, Trustworthy Autonomous Systems
@inproceedings{soton446459,
title = {Responsibility ascription in trustworthy autonomous systems},
author = {Vahid Yazdanpanah and Enrico H Gerding and Sebastian Stein and Mehdi Dastani and Catholijn M Jonker and Timothy J Norman and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/446459/},
year = {2021},
date = {2021-02-01},
booktitle = {Embedding AI in Society (18/02/21 - 19/02/21)},
abstract = {To develop and effectively deploy Trustworthy Autonomous Systems (TAS), we face various social, technological, legal, and ethical challenges in which different notions of responsibility can play a key role. In this work, we elaborate on these challenges, discuss research gaps, and show how the multidimensional notion of responsibility can play a key role to bridge them. We argue that TAS requires operational tools to represent and reason about the responsibilities of humans as well as AI agents.},
keywords = {Multiagent Systems, Reliable AI, Responsibility Reasoning, Trustworthy Autonomous Systems},
pubstate = {published},
tppubtype = {inproceedings}
}
To develop and effectively deploy Trustworthy Autonomous Systems (TAS), we face various social, technological, legal, and ethical challenges in which different notions of responsibility can play a key role. In this work, we elaborate on these challenges, discuss research gaps, and show how the multidimensional notion of responsibility can play a key role to bridge them. We argue that TAS requires operational tools to represent and reason about the responsibilities of humans as well as AI agents.
Beal, Ryan James; Middleton, Stuart; Norman, Timothy; Ramchurn, Sarvapali
Combining machine learning and human experts to predict match outcomes in football: A baseline model Proceedings Article
In: The Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (02/02/21 - 09/02/21), 2021.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Natural Language Processing
@inproceedings{soton445607,
title = {Combining machine learning and human experts to predict match outcomes in football: A baseline model},
author = {Ryan James Beal and Stuart Middleton and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/445607/},
year = {2021},
date = {2021-02-01},
booktitle = {The Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (02/02/21 - 09/02/21)},
abstract = {In this paper, we present a new application-focused benchmark dataset and results from a set of baseline Natural Language Processing and Machine Learning models for prediction of match outcomes for games of football (soccer). By doing so we give a baseline for the prediction accuracy that can be achieved exploiting both statistical match data and contextual articles from human sports journalists. Our dataset is focuses on a representative time-period over 6 seasons of the English Premier League, and includes newspaper match previews from The Guardian. The models presented in this paper achieve an accuracy of 63.18% showing a 6.9% boost on the traditional statistical methods.},
keywords = {Artificial Intelligence, Natural Language Processing},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we present a new application-focused benchmark dataset and results from a set of baseline Natural Language Processing and Machine Learning models for prediction of match outcomes for games of football (soccer). By doing so we give a baseline for the prediction accuracy that can be achieved exploiting both statistical match data and contextual articles from human sports journalists. Our dataset is focuses on a representative time-period over 6 seasons of the English Premier League, and includes newspaper match previews from The Guardian. The models presented in this paper achieve an accuracy of 63.18% showing a 6.9% boost on the traditional statistical methods.
Lhopital, Sacha; Aknine, Samir; Ramchurn, Sarvapali; Thavonekham, Vincent; Vu, Huan
Decentralised control of intelligent devices: a healthcare facility study Proceedings Article
In: Bassiliades, Nick; Chalkiadakis, Georgios; Jonge, Dave (Ed.): Multi-Agent Systems and Agreement Technologies - 17th European Conference, EUMAS 2020, and 7th International Conference, AT 2020, Revised Selected Papers, pp. 20–36, Springer, 2021.
Abstract | Links | BibTeX | Tags: DCOP, DPOP, Healthcare, IoT
@inproceedings{soton447983,
title = {Decentralised control of intelligent devices: a healthcare facility study},
author = {Sacha Lhopital and Samir Aknine and Sarvapali Ramchurn and Vincent Thavonekham and Huan Vu},
editor = {Nick Bassiliades and Georgios Chalkiadakis and Dave Jonge},
url = {https://eprints.soton.ac.uk/447983/},
year = {2021},
date = {2021-01-01},
booktitle = {Multi-Agent Systems and Agreement Technologies - 17th European Conference, EUMAS 2020, and 7th International Conference, AT 2020, Revised Selected Papers},
volume = {12520 LNAI},
pages = {20–36},
publisher = {Springer},
abstract = {ensuremath<pensuremath>We present a novel approach to the management of notifications from devices in a healthcare setting. We employ a distributed constraint optimisation (DCOP) approach to the delivery of notification for healthcare assistants that aims to preserve the privacy of patients while reducing the intrusiveness of such notifications. Our approach reduces the workload of the assistants and improves patient safety by automating task allocation while ensuring high priority needs are addressed in a timely manner. We propose and evaluate several DCOP models both in simulation and in real-world deployments. Our models are shown to be efficient both in terms of computation and communication costs.ensuremath</pensuremath>},
keywords = {DCOP, DPOP, Healthcare, IoT},
pubstate = {published},
tppubtype = {inproceedings}
}
ensuremath<pensuremath>We present a novel approach to the management of notifications from devices in a healthcare setting. We employ a distributed constraint optimisation (DCOP) approach to the delivery of notification for healthcare assistants that aims to preserve the privacy of patients while reducing the intrusiveness of such notifications. Our approach reduces the workload of the assistants and improves patient safety by automating task allocation while ensuring high priority needs are addressed in a timely manner. We propose and evaluate several DCOP models both in simulation and in real-world deployments. Our models are shown to be efficient both in terms of computation and communication costs.ensuremath</pensuremath>
Ryan, James Beal; Chalkiadakis, Georgios; Norman, Timothy; Ramchurn, Sarvapali
Optimising long-term outcomes using real-world fluent objectives: an application to football Proceedings Article
In: 20th International Conference on Autonomous Agents and Multiagent Systems (03/05/21 - 07/05/21), pp. 196–204, 2021.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton449655,
title = {Optimising long-term outcomes using real-world fluent objectives: an application to football},
author = {James Beal Ryan and Georgios Chalkiadakis and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449655/},
year = {2021},
date = {2021-01-01},
booktitle = {20th International Conference on Autonomous Agents and Multiagent Systems (03/05/21 - 07/05/21)},
pages = {196–204},
abstract = {In this paper, we present a novel approach for optimising long-term tactical and strategic decision-making in football (soccer) by encapsulating events in a league environment across a given time frame. ensuremath<br/ensuremath>We model the teams' objectives for a season and track how these evolve as games unfold to give a fluent objective that can aid in decision-making games. We develop Markov chain Monte Carlo and deep learning-based algorithms that make use of the fluent objectives in order to learn from prior games and other games in the environment and increase the teams? long-term performance. Simulations of our approach using real-world datasets from 760 matches shows that by using optimised tactics with our fluent objective and prior games, we can on average increase teams mean expected finishing distribution in the league by up to 35.6%.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we present a novel approach for optimising long-term tactical and strategic decision-making in football (soccer) by encapsulating events in a league environment across a given time frame. ensuremath<br/ensuremath>We model the teams' objectives for a season and track how these evolve as games unfold to give a fluent objective that can aid in decision-making games. We develop Markov chain Monte Carlo and deep learning-based algorithms that make use of the fluent objectives in order to learn from prior games and other games in the environment and increase the teams? long-term performance. Simulations of our approach using real-world datasets from 760 matches shows that by using optimised tactics with our fluent objective and prior games, we can on average increase teams mean expected finishing distribution in the league by up to 35.6%.
Merhej, Charbel; Ryan, James Beal; Matthews, Tim; Ramchurn, Sarvapali
What happened next? Using deep learning to value defensive actions in football event-data Proceedings Article
In: KDD 2021 (14/08/21 - 18/08/21), pp. 3394–3403, 2021.
Abstract | Links | BibTeX | Tags: applied machine learning, deep learning, defensive actions, football, neural networks, sports analytics
@inproceedings{soton449656,
title = {What happened next? Using deep learning to value defensive actions in football event-data},
author = {Charbel Merhej and James Beal Ryan and Tim Matthews and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449656/},
year = {2021},
date = {2021-01-01},
booktitle = {KDD 2021 (14/08/21 - 18/08/21)},
pages = {3394–3403},
abstract = {Objectively quantifying the value of player actions in football (soccer) is a challenging problem. To date, studies in football analytics have mainly focused on the attacking side of the game, while there has been less work on event-driven metrics for valuing defensive actions (e.g., tackles and interceptions). Therefore in this paper, we use deep learning techniques to define a novel metric that values such defensive actions by studying the threat of passages of play that preceded them. By doing so, we are able to value defensive actions based on what they prevented from happening in the game. Our Defensive Action Expected Threat (DAxT) model has been validated using real-world event-data from the 2017/2018 and 2018/2019 English Premier League seasons, and we combine our model outputs with additional features to derive an overall rating of defensive ability for players. Overall, we find that our model is able to predict the impact of defensive actions allowing us to better value defenders using event-data.},
keywords = {applied machine learning, deep learning, defensive actions, football, neural networks, sports analytics},
pubstate = {published},
tppubtype = {inproceedings}
}
Objectively quantifying the value of player actions in football (soccer) is a challenging problem. To date, studies in football analytics have mainly focused on the attacking side of the game, while there has been less work on event-driven metrics for valuing defensive actions (e.g., tackles and interceptions). Therefore in this paper, we use deep learning techniques to define a novel metric that values such defensive actions by studying the threat of passages of play that preceded them. By doing so, we are able to value defensive actions based on what they prevented from happening in the game. Our Defensive Action Expected Threat (DAxT) model has been validated using real-world event-data from the 2017/2018 and 2018/2019 English Premier League seasons, and we combine our model outputs with additional features to derive an overall rating of defensive ability for players. Overall, we find that our model is able to predict the impact of defensive actions allowing us to better value defenders using event-data.
2020
Beal, Ryan James; Norman, Timothy; Ramchurn, Sarvapali
Optimising daily fantasy sports teams with artificial intelligence Journal Article
In: International Journal of Computer Science in Sport, vol. 19, no. 2, 2020.
Abstract | Links | BibTeX | Tags:
@article{soton445995,
title = {Optimising daily fantasy sports teams with artificial intelligence},
author = {Ryan James Beal and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/445995/},
year = {2020},
date = {2020-12-01},
journal = {International Journal of Computer Science in Sport},
volume = {19},
number = {2},
abstract = {This paper outlines a novel approach to optimising teams for Daily Fantasy Sports (DFS) contests. To this end, we propose a number of new models and algorithms to solve the team formation problems posed by DFS. Specifically, we focus on the National Football League (NFL) and predict the performance of real-world players to form the optimal fantasy team using mixed-integer programming. We test our solutions using real-world data-sets from across four seasons (2014-2017). We highlight the advantage that can be gained from using our machine-based methods and show that our solutions outperform existing benchmarks, turning a profit in up to 81.3% of DFS game-weeks over a season},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper outlines a novel approach to optimising teams for Daily Fantasy Sports (DFS) contests. To this end, we propose a number of new models and algorithms to solve the team formation problems posed by DFS. Specifically, we focus on the National Football League (NFL) and predict the performance of real-world players to form the optimal fantasy team using mixed-integer programming. We test our solutions using real-world data-sets from across four seasons (2014-2017). We highlight the advantage that can be gained from using our machine-based methods and show that our solutions outperform existing benchmarks, turning a profit in up to 81.3% of DFS game-weeks over a season
Beal, Ryan James; Norman, Timothy; Ramchurn, Sarvapali
A critical comparison of machine learning classifiers to predict match outcomes in the NFL Journal Article
In: International Journal of Computer Science in Sport, vol. 19, no. 2, 2020.
Abstract | Links | BibTeX | Tags:
@article{soton446078,
title = {A critical comparison of machine learning classifiers to predict match outcomes in the NFL},
author = {Ryan James Beal and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/446078/},
year = {2020},
date = {2020-12-01},
journal = {International Journal of Computer Science in Sport},
volume = {19},
number = {2},
abstract = {In this paper, we critically evaluate the performance of nine machine learning classification techniques when applied to the match outcome prediction problem presented by American Football. Specifically, we implement and test nine techniques using real-world datasets of 1280 games over 5 seasons from the National Football League (NFL). We test the nine different classifier techniques using a total of 42 features for each team and we find that the best performing algorithms are able to improve one previous published works. The algoriothms achieve an accuracy of between 44.64% for a Guassian Process classifier to 67.53% with a Na"ive Bayes classifer. We also test each classifier on a year by year basis and compare our results to those of the bookmakers and other leading academic papers.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper, we critically evaluate the performance of nine machine learning classification techniques when applied to the match outcome prediction problem presented by American Football. Specifically, we implement and test nine techniques using real-world datasets of 1280 games over 5 seasons from the National Football League (NFL). We test the nine different classifier techniques using a total of 42 features for each team and we find that the best performing algorithms are able to improve one previous published works. The algoriothms achieve an accuracy of between 44.64% for a Guassian Process classifier to 67.53% with a Na"ive Bayes classifer. We also test each classifier on a year by year basis and compare our results to those of the bookmakers and other leading academic papers.
Rigas, Emmanouil S; Gerding, Enrico; Stein, Sebastian; Ramchurn, Sarvapali D; Bassiliades, Nick
Mechanism design for efficient allocation of electric vehicles to charging stations Proceedings Article
In: SETN 2020: 11th Hellenic Conference on Artificial Intelligence, pp. 10–15, 2020.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton446412,
title = {Mechanism design for efficient allocation of electric vehicles to charging stations},
author = {Emmanouil S Rigas and Enrico Gerding and Sebastian Stein and Sarvapali D Ramchurn and Nick Bassiliades},
url = {https://eprints.soton.ac.uk/446412/},
year = {2020},
date = {2020-09-01},
booktitle = {SETN 2020: 11th Hellenic Conference on Artificial Intelligence},
pages = {10–15},
abstract = {The electrification of transport can significantly reduce CO2 emissions and their negative impact on the environment. In this paper, we study the problem of allocating Electric Vehicles (EVs) to charging stations and scheduling their charging. We develop an offline solution that treats EV users as self-interested agents that aim to maximise their profit and minimise the impact on their schedule. We formulate the problem of the optimal EV to charging station allocation as a Mixed Integer Programming (MIP) one and we propose two pricing mechanisms: A fixed-price one, and another that is based on the well known Vickrey-Clark-Groves (VCG) mechanism. We observe that the VCG mechanism services on average 1.5% more EVs than the fixed-price one. In addition, when the stations get congested, VCG leads to higher prices for the EVs and higher profit for the stations, but lower utility for the EVs. However, the VCG mechanism guarantees truthful reporting of the EVs? preferences.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The electrification of transport can significantly reduce CO2 emissions and their negative impact on the environment. In this paper, we study the problem of allocating Electric Vehicles (EVs) to charging stations and scheduling their charging. We develop an offline solution that treats EV users as self-interested agents that aim to maximise their profit and minimise the impact on their schedule. We formulate the problem of the optimal EV to charging station allocation as a Mixed Integer Programming (MIP) one and we propose two pricing mechanisms: A fixed-price one, and another that is based on the well known Vickrey-Clark-Groves (VCG) mechanism. We observe that the VCG mechanism services on average 1.5% more EVs than the fixed-price one. In addition, when the stations get congested, VCG leads to higher prices for the EVs and higher profit for the stations, but lower utility for the EVs. However, the VCG mechanism guarantees truthful reporting of the EVs? preferences.
Oluwasuji, Olabambo Ifeoluwa; Malik, Obaid; Zhang, Jie; Ramchurn, Sarvapali Dyanand
Solving the fair electric load shedding problem in developing countries Journal Article
In: Autonomous Agents and Multi-Agent Systems, vol. 34, no. 1, pp. 12, 2020.
BibTeX | Tags:
@article{oluwasuji2020solving,
title = {Solving the fair electric load shedding problem in developing countries},
author = {Olabambo Ifeoluwa Oluwasuji and Obaid Malik and Jie Zhang and Sarvapali Dyanand Ramchurn},
year = {2020},
date = {2020-01-01},
journal = {Autonomous Agents and Multi-Agent Systems},
volume = {34},
number = {1},
pages = {12},
publisher = {Springer},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Oluwasuji, Olabambo Ifeoluwa; Malik, Obaid; Zhang, Jie; Ramchurn, Sarvapali Dyanand
Solving the fair electric load shedding problem in developing countries Journal Article
In: Auton. Agents Multi Agent Syst., vol. 34, no. 1, pp. 12, 2020.
@article{DBLP:journals/aamas/OluwasujiMZR20,
title = {Solving the fair electric load shedding problem in developing countries},
author = {Olabambo Ifeoluwa Oluwasuji and Obaid Malik and Jie Zhang and Sarvapali Dyanand Ramchurn},
url = {https://doi.org/10.1007/s10458-019-09428-8},
doi = {10.1007/s10458-019-09428-8},
year = {2020},
date = {2020-01-01},
journal = {Auton. Agents Multi Agent Syst.},
volume = {34},
number = {1},
pages = {12},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
1.
Deshmukh, Jayati; Liang, Zijie; Yazdanpanah, Vahid; Stein, Sebastian; Ramchurn, Sarvalpali D.
Serious games for ethical preference elicitation Proceedings Article
In: AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25), 2025.
@inproceedings{soton498743,
title = {Serious games for ethical preference elicitation},
author = {Jayati Deshmukh and Zijie Liang and Vahid Yazdanpanah and Sebastian Stein and Sarvalpali D. Ramchurn},
url = {https://eprints.soton.ac.uk/498743/},
year = {2025},
date = {2025-05-01},
booktitle = {AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25)},
abstract = {Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.
2.
Thavanesan, Navamayooran; Farahi, Arya; Parfitt, Charlotte; Belkhatir, Zehor; Azim, Tayyaba; Vallejos, Elvira Perez; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Insights from explainable AI in oesophageal cancer team decisions Journal Article
In: Computers in Biology and Medicine, vol. 180, 2024, (For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.).
@article{soton493238,
title = {Insights from explainable AI in oesophageal cancer team decisions},
author = {Navamayooran Thavanesan and Arya Farahi and Charlotte Parfitt and Zehor Belkhatir and Tayyaba Azim and Elvira Perez Vallejos and Zoë Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/493238/},
year = {2024},
date = {2024-08-01},
journal = {Computers in Biology and Medicine},
volume = {180},
abstract = {ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>},
note = {For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>
3.
Naiseh, Mohammad; Webb, Catherine; Underwood, Tim; Ramchurn, Gopal; Walters, Zoe; Thavanesan, Navamayooran; Vigneswaran, Ganesh
XAI for group-AI interaction: towards collaborative and inclusive explanations Proceedings Article
In: Longo, Luca; Liu, Weiru; Montavon, Gregoire (Ed.): Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024), pp. 249–256, CEUR Workshop Proceedings, 2024.
@inproceedings{soton497829,
title = {XAI for group-AI interaction: towards collaborative and inclusive explanations},
author = {Mohammad Naiseh and Catherine Webb and Tim Underwood and Gopal Ramchurn and Zoe Walters and Navamayooran Thavanesan and Ganesh Vigneswaran},
editor = {Luca Longo and Weiru Liu and Gregoire Montavon},
url = {https://eprints.soton.ac.uk/497829/},
year = {2024},
date = {2024-07-01},
booktitle = {Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024)},
volume = {3793},
pages = {249–256},
publisher = {CEUR Workshop Proceedings},
abstract = {ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>
4.
Early, Joseph Arthur
Interpretable multiple instance learning PhD Thesis
University of Southampton, 2024.
@phdthesis{soton490767,
title = {Interpretable multiple instance learning},
author = {Joseph Arthur Early},
url = {https://eprints.soton.ac.uk/490767/},
year = {2024},
date = {2024-06-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>
5.
Kiden, Sarah; Stahl, Bernd; Townsend, Beverley; Maple, Carsten; Vincent, Charles; Sampson, Fraser; Gilbert, Geoff; Smith, Helen; Deshmukh, Jayati; Ross, Jen; Williams, Jennifer; Rincon, Jesus Martinez; Lisinska, Justyna; O?Shea, Karen; Abreu, Márjory Da Costa; Bencomo, Nelly; Deb, Oishi; Winter, Peter; Li, Phoebe; Torr, Philip; Lau, Pin Lean; Iniesta, Raquel; Ramchurn, Gopal; Stein, Sebastian; Yazdanpanah, Vahid
Responsible AI governance: A response to UN interim report on governing AI for humanity Technical Report
no. 10.5258/SOTON/PP0057, 2024.
@techreport{soton488908,
title = {Responsible AI governance: A response to UN interim report on governing AI for humanity},
author = {Sarah Kiden and Bernd Stahl and Beverley Townsend and Carsten Maple and Charles Vincent and Fraser Sampson and Geoff Gilbert and Helen Smith and Jayati Deshmukh and Jen Ross and Jennifer Williams and Jesus Martinez Rincon and Justyna Lisinska and Karen O?Shea and Márjory Da Costa Abreu and Nelly Bencomo and Oishi Deb and Peter Winter and Phoebe Li and Philip Torr and Pin Lean Lau and Raquel Iniesta and Gopal Ramchurn and Sebastian Stein and Vahid Yazdanpanah},
url = {https://eprints.soton.ac.uk/488908/},
year = {2024},
date = {2024-03-01},
number = {10.5258/SOTON/PP0057},
publisher = {Public Policy, University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
6.
Abioye, Ayodeji O.; Hunt, William; Gu, Yue; Schneiders, Eike; Naiseh, Mohammad; Fischer, Joel E.; Ramchurn, Sarvapali D.; Soorati, Mohammad D.; Archibald, Blair; Sevegnani, Michele
The effect of predictive formal modelling at runtime on performance in human-swarm interaction Proceedings Article
In: HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction, pp. 172?176, Association for Computing Machinery, 2024, (Publisher Copyright: © 2024 Copyright held by the owner/author(s)).
@inproceedings{soton488273,
title = {The effect of predictive formal modelling at runtime on performance in human-swarm interaction},
author = {Ayodeji O. Abioye and William Hunt and Yue Gu and Eike Schneiders and Mohammad Naiseh and Joel E. Fischer and Sarvapali D. Ramchurn and Mohammad D. Soorati and Blair Archibald and Michele Sevegnani},
url = {https://eprints.soton.ac.uk/488273/},
year = {2024},
date = {2024-03-01},
booktitle = {HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction},
pages = {172?176},
publisher = {Association for Computing Machinery},
abstract = {Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.},
note = {Publisher Copyright:
© 2024 Copyright held by the owner/author(s)},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.
7.
Soorati, Mohammad D.; Naiseh, Mohammad; Hunt, William; Parnell, Katie; Clark, Jediah; Ramchurn, Sarvapali D.
Enabling trustworthiness in human-swarm systems through a digital twin Book Section
In: Dasgupta, Prithviraj; Llinas, James; Gillespie, Tony; Fouse, Scott; Lawless, William; Mittu, Ranjeev; Sofge, Donlad (Ed.): Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams, pp. 93–125, Academic Press, 2024, (Publisher Copyright: © 2024 Elsevier Inc. All rights reserved.).
@incollection{soton491769,
title = {Enabling trustworthiness in human-swarm systems through a digital twin},
author = {Mohammad D. Soorati and Mohammad Naiseh and William Hunt and Katie Parnell and Jediah Clark and Sarvapali D. Ramchurn},
editor = {Prithviraj Dasgupta and James Llinas and Tony Gillespie and Scott Fouse and William Lawless and Ranjeev Mittu and Donlad Sofge},
url = {https://eprints.soton.ac.uk/491769/},
year = {2024},
date = {2024-02-01},
booktitle = {Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams},
pages = {93–125},
publisher = {Academic Press},
abstract = {Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.},
note = {Publisher Copyright:
© 2024 Elsevier Inc. All rights reserved.},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.
8.
Thavanesan, Navamayooran; Parfitt, Charlotte; Bodala, Indu; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy; Vigneswaran, Ganesh
Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction Miscellaneous
2024.
@misc{soton497828,
title = {Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction},
author = {Navamayooran Thavanesan and Charlotte Parfitt and Indu Bodala and Zoë Walters and Sarvapali Ramchurn and Timothy Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/497828/},
year = {2024},
date = {2024-01-01},
journal = {European Journal of Surgical Oncology},
volume = {50},
number = {1},
abstract = {Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035
9.
Kelly, Thomas Graham; Soorati, Mohammad; Zauner, Klaus-Peter; Ramchurn, Gopal; Tarapore, Danesh
Trade-offs of dynamic control structure in human-swarm systems Proceedings Article
In: The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024, 2024.
@inproceedings{soton492838,
title = {Trade-offs of dynamic control structure in human-swarm systems},
author = {Thomas Graham Kelly and Mohammad Soorati and Klaus-Peter Zauner and Gopal Ramchurn and Danesh Tarapore},
url = {https://eprints.soton.ac.uk/492838/},
year = {2024},
date = {2024-01-01},
booktitle = {The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024},
abstract = {Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.
10.
Early, Joseph; Deweese, Ying-Jung Chen; Evers, Christine; Ramchurn, Sarvapali
Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation Journal Article
In: Environmental Data Science, vol. 2, pp. 18, 2023.
@article{soton490766,
title = {Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation},
author = {Joseph Early and Ying-Jung Chen Deweese and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/490766/},
year = {2023},
date = {2023-12-01},
journal = {Environmental Data Science},
volume = {2},
pages = {18},
abstract = {Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.
11.
Rigley, Eryn; Bentley, Caitlin; Krook, Joshua; Ramchurn, Gopal
Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries Journal Article
In: Global Policy, 2023, (Funding Information: This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.).
@article{soton485727,
title = {Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries},
author = {Eryn Rigley and Caitlin Bentley and Joshua Krook and Gopal Ramchurn},
url = {https://eprints.soton.ac.uk/485727/},
year = {2023},
date = {2023-12-01},
journal = {Global Policy},
abstract = {ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>},
note = {Funding Information:
This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>
12.
Singh, Lokesh; Ramchurn, Gopal
The effect of automated agents on individual performance under induced stress Proceedings Article
In: Kalra, Jay (Ed.): Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition), pp. 118–127, AHFE International, 2023.
@inproceedings{soton485655,
title = {The effect of automated agents on individual performance under induced stress},
author = {Lokesh Singh and Gopal Ramchurn},
editor = {Jay Kalra},
url = {https://eprints.soton.ac.uk/485655/},
year = {2023},
date = {2023-11-01},
booktitle = {Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition)},
pages = {118–127},
publisher = {AHFE International},
abstract = {Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.
13.
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, vol. 49, no. 11, 2023, (Publisher Copyright: © 2023 The Author(s)).
@article{soton479497b,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-11-01},
journal = {European Journal of Surgical Oncology},
volume = {49},
number = {11},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
© 2023 The Author(s)},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
14.
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub Response Miscellaneous
2023.
@misc{soton481740,
title = {AI large language models inquiry: TASHub Response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
15.
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub response Miscellaneous
2023.
@misc{soton481740b,
title = {AI large language models inquiry: TASHub response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
16.
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, 2023, (Publisher Copyright: copyright 2023 The Author(s)).
@article{soton479497,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-07-01},
journal = {European Journal of Surgical Oncology},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
copyright 2023 The Author(s)},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
17.
Abioye, Ayodeji
University of Southampton, 2023.
@phdthesis{soton479472,
title = {Multimodal speech and visual gesture control interface technique for small unmanned multirotor aircraft},
author = {Ayodeji Abioye},
url = {https://eprints.soton.ac.uk/479472/},
year = {2023},
date = {2023-07-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>
18.
Abioye, Ayodeji; Naiseh, Mohammad; Hunt, William; Clark, Jediah R; Ramchurn, Sarvapali D; Soorati, Mohammad
The effect of data visualisation quality and task density on human-swarm interaction Proceedings Article
In: Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), IEEE, 2023.
@inproceedings{soton479970,
title = {The effect of data visualisation quality and task density on human-swarm interaction},
author = {Ayodeji Abioye and Mohammad Naiseh and William Hunt and Jediah R Clark and Sarvapali D Ramchurn and Mohammad Soorati},
url = {https://eprints.soton.ac.uk/479970/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
booktitle = {Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
publisher = {IEEE},
abstract = {Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.
19.
Krook, Joshua; McAuley, Derek; Anderson, Stuart; Downer, John; Winter, Peter; Ramchurn, Sarvapali D
AI Foundation Models: initial review, CMA Consultation, TAS Hub Response Miscellaneous
2023.
@misc{soton477553,
title = {AI Foundation Models: initial review, CMA Consultation, TAS Hub Response},
author = {Joshua Krook and Derek McAuley and Stuart Anderson and John Downer and Peter Winter and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477553/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
publisher = {University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
20.
Krook, Joshua; Downer, John; Winter, Peter; Williams, Jennifer; Ives, Jonathan; Bratu, Roxana; Sheir, Stephanie; Williams, Robin; Anderson, Stuart; Li, Phoebe; Ramamoorthy, Subramanian; Ramchurn, Sarvapali
AI regulation: a pro-innovation approach ? policy proposals: TASHub Response Miscellaneous
2023.
@misc{soton478329,
title = {AI regulation: a pro-innovation approach ? policy proposals: TASHub Response},
author = {Joshua Krook and John Downer and Peter Winter and Jennifer Williams and Jonathan Ives and Roxana Bratu and Stephanie Sheir and Robin Williams and Stuart Anderson and Phoebe Li and Subramanian Ramamoorthy and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478329/},
year = {2023},
date = {2023-06-01},
publisher = {University of Southampton},
abstract = {Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence
2025
Deshmukh, Jayati; Liang, Zijie; Yazdanpanah, Vahid; Stein, Sebastian; Ramchurn, Sarvalpali D.
Serious games for ethical preference elicitation Proceedings Article
In: AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25), 2025.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton498743,
title = {Serious games for ethical preference elicitation},
author = {Jayati Deshmukh and Zijie Liang and Vahid Yazdanpanah and Sebastian Stein and Sarvalpali D. Ramchurn},
url = {https://eprints.soton.ac.uk/498743/},
year = {2025},
date = {2025-05-01},
booktitle = {AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25)},
abstract = {Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.
2024
Thavanesan, Navamayooran; Farahi, Arya; Parfitt, Charlotte; Belkhatir, Zehor; Azim, Tayyaba; Vallejos, Elvira Perez; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Insights from explainable AI in oesophageal cancer team decisions Journal Article
In: Computers in Biology and Medicine, vol. 180, 2024, (For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.).
Abstract | Links | BibTeX | Tags: Decision-making, machine learning, Multidisciplinary teams, Oesophageal cancer
@article{soton493238,
title = {Insights from explainable AI in oesophageal cancer team decisions},
author = {Navamayooran Thavanesan and Arya Farahi and Charlotte Parfitt and Zehor Belkhatir and Tayyaba Azim and Elvira Perez Vallejos and Zoë Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/493238/},
year = {2024},
date = {2024-08-01},
journal = {Computers in Biology and Medicine},
volume = {180},
abstract = {ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>},
note = {For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.},
keywords = {Decision-making, machine learning, Multidisciplinary teams, Oesophageal cancer},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>
Naiseh, Mohammad; Webb, Catherine; Underwood, Tim; Ramchurn, Gopal; Walters, Zoe; Thavanesan, Navamayooran; Vigneswaran, Ganesh
XAI for group-AI interaction: towards collaborative and inclusive explanations Proceedings Article
In: Longo, Luca; Liu, Weiru; Montavon, Gregoire (Ed.): Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024), pp. 249–256, CEUR Workshop Proceedings, 2024.
Abstract | Links | BibTeX | Tags: Explainable AI, Group-AI Interaction, Interaction Design
@inproceedings{soton497829,
title = {XAI for group-AI interaction: towards collaborative and inclusive explanations},
author = {Mohammad Naiseh and Catherine Webb and Tim Underwood and Gopal Ramchurn and Zoe Walters and Navamayooran Thavanesan and Ganesh Vigneswaran},
editor = {Luca Longo and Weiru Liu and Gregoire Montavon},
url = {https://eprints.soton.ac.uk/497829/},
year = {2024},
date = {2024-07-01},
booktitle = {Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024)},
volume = {3793},
pages = {249–256},
publisher = {CEUR Workshop Proceedings},
abstract = {ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>},
keywords = {Explainable AI, Group-AI Interaction, Interaction Design},
pubstate = {published},
tppubtype = {inproceedings}
}
ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>
Early, Joseph Arthur
Interpretable multiple instance learning PhD Thesis
University of Southampton, 2024.
Abstract | Links | BibTeX | Tags:
@phdthesis{soton490767,
title = {Interpretable multiple instance learning},
author = {Joseph Arthur Early},
url = {https://eprints.soton.ac.uk/490767/},
year = {2024},
date = {2024-06-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>
Kiden, Sarah; Stahl, Bernd; Townsend, Beverley; Maple, Carsten; Vincent, Charles; Sampson, Fraser; Gilbert, Geoff; Smith, Helen; Deshmukh, Jayati; Ross, Jen; Williams, Jennifer; Rincon, Jesus Martinez; Lisinska, Justyna; O?Shea, Karen; Abreu, Márjory Da Costa; Bencomo, Nelly; Deb, Oishi; Winter, Peter; Li, Phoebe; Torr, Philip; Lau, Pin Lean; Iniesta, Raquel; Ramchurn, Gopal; Stein, Sebastian; Yazdanpanah, Vahid
Responsible AI governance: A response to UN interim report on governing AI for humanity Technical Report
no. 10.5258/SOTON/PP0057, 2024.
@techreport{soton488908,
title = {Responsible AI governance: A response to UN interim report on governing AI for humanity},
author = {Sarah Kiden and Bernd Stahl and Beverley Townsend and Carsten Maple and Charles Vincent and Fraser Sampson and Geoff Gilbert and Helen Smith and Jayati Deshmukh and Jen Ross and Jennifer Williams and Jesus Martinez Rincon and Justyna Lisinska and Karen O?Shea and Márjory Da Costa Abreu and Nelly Bencomo and Oishi Deb and Peter Winter and Phoebe Li and Philip Torr and Pin Lean Lau and Raquel Iniesta and Gopal Ramchurn and Sebastian Stein and Vahid Yazdanpanah},
url = {https://eprints.soton.ac.uk/488908/},
year = {2024},
date = {2024-03-01},
number = {10.5258/SOTON/PP0057},
publisher = {Public Policy, University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Abioye, Ayodeji O.; Hunt, William; Gu, Yue; Schneiders, Eike; Naiseh, Mohammad; Fischer, Joel E.; Ramchurn, Sarvapali D.; Soorati, Mohammad D.; Archibald, Blair; Sevegnani, Michele
The effect of predictive formal modelling at runtime on performance in human-swarm interaction Proceedings Article
In: HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction, pp. 172?176, Association for Computing Machinery, 2024, (Publisher Copyright: © 2024 Copyright held by the owner/author(s)).
Abstract | Links | BibTeX | Tags: Human-Robot Interaction (HRI), Human-Swarm Interaction (HSI), Predictive Formal Modelling (PFM), Task Performance
@inproceedings{soton488273,
title = {The effect of predictive formal modelling at runtime on performance in human-swarm interaction},
author = {Ayodeji O. Abioye and William Hunt and Yue Gu and Eike Schneiders and Mohammad Naiseh and Joel E. Fischer and Sarvapali D. Ramchurn and Mohammad D. Soorati and Blair Archibald and Michele Sevegnani},
url = {https://eprints.soton.ac.uk/488273/},
year = {2024},
date = {2024-03-01},
booktitle = {HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction},
pages = {172?176},
publisher = {Association for Computing Machinery},
abstract = {Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.},
note = {Publisher Copyright:
© 2024 Copyright held by the owner/author(s)},
keywords = {Human-Robot Interaction (HRI), Human-Swarm Interaction (HSI), Predictive Formal Modelling (PFM), Task Performance},
pubstate = {published},
tppubtype = {inproceedings}
}
Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.
Soorati, Mohammad D.; Naiseh, Mohammad; Hunt, William; Parnell, Katie; Clark, Jediah; Ramchurn, Sarvapali D.
Enabling trustworthiness in human-swarm systems through a digital twin Book Section
In: Dasgupta, Prithviraj; Llinas, James; Gillespie, Tony; Fouse, Scott; Lawless, William; Mittu, Ranjeev; Sofge, Donlad (Ed.): Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams, pp. 93–125, Academic Press, 2024, (Publisher Copyright: © 2024 Elsevier Inc. All rights reserved.).
Abstract | Links | BibTeX | Tags: Digital twin, Explainability, Human-swarm interaction, Trustworthy Autonomous Systems, User-centered design
@incollection{soton491769,
title = {Enabling trustworthiness in human-swarm systems through a digital twin},
author = {Mohammad D. Soorati and Mohammad Naiseh and William Hunt and Katie Parnell and Jediah Clark and Sarvapali D. Ramchurn},
editor = {Prithviraj Dasgupta and James Llinas and Tony Gillespie and Scott Fouse and William Lawless and Ranjeev Mittu and Donlad Sofge},
url = {https://eprints.soton.ac.uk/491769/},
year = {2024},
date = {2024-02-01},
booktitle = {Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams},
pages = {93–125},
publisher = {Academic Press},
abstract = {Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.},
note = {Publisher Copyright:
© 2024 Elsevier Inc. All rights reserved.},
keywords = {Digital twin, Explainability, Human-swarm interaction, Trustworthy Autonomous Systems, User-centered design},
pubstate = {published},
tppubtype = {incollection}
}
Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.
Thavanesan, Navamayooran; Parfitt, Charlotte; Bodala, Indu; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy; Vigneswaran, Ganesh
Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction Miscellaneous
2024.
Abstract | Links | BibTeX | Tags:
@misc{soton497828,
title = {Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction},
author = {Navamayooran Thavanesan and Charlotte Parfitt and Indu Bodala and Zoë Walters and Sarvapali Ramchurn and Timothy Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/497828/},
year = {2024},
date = {2024-01-01},
journal = {European Journal of Surgical Oncology},
volume = {50},
number = {1},
abstract = {Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035
Kelly, Thomas Graham; Soorati, Mohammad; Zauner, Klaus-Peter; Ramchurn, Gopal; Tarapore, Danesh
Trade-offs of dynamic control structure in human-swarm systems Proceedings Article
In: The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024, 2024.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton492838,
title = {Trade-offs of dynamic control structure in human-swarm systems},
author = {Thomas Graham Kelly and Mohammad Soorati and Klaus-Peter Zauner and Gopal Ramchurn and Danesh Tarapore},
url = {https://eprints.soton.ac.uk/492838/},
year = {2024},
date = {2024-01-01},
booktitle = {The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024},
abstract = {Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.
2023
Early, Joseph; Deweese, Ying-Jung Chen; Evers, Christine; Ramchurn, Sarvapali
Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation Journal Article
In: Environmental Data Science, vol. 2, pp. 18, 2023.
Abstract | Links | BibTeX | Tags:
@article{soton490766,
title = {Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation},
author = {Joseph Early and Ying-Jung Chen Deweese and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/490766/},
year = {2023},
date = {2023-12-01},
journal = {Environmental Data Science},
volume = {2},
pages = {18},
abstract = {Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.
Rigley, Eryn; Bentley, Caitlin; Krook, Joshua; Ramchurn, Gopal
Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries Journal Article
In: Global Policy, 2023, (Funding Information: This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.).
Abstract | Links | BibTeX | Tags:
@article{soton485727,
title = {Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries},
author = {Eryn Rigley and Caitlin Bentley and Joshua Krook and Gopal Ramchurn},
url = {https://eprints.soton.ac.uk/485727/},
year = {2023},
date = {2023-12-01},
journal = {Global Policy},
abstract = {ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>},
note = {Funding Information:
This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>
Singh, Lokesh; Ramchurn, Gopal
The effect of automated agents on individual performance under induced stress Proceedings Article
In: Kalra, Jay (Ed.): Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition), pp. 118–127, AHFE International, 2023.
Abstract | Links | BibTeX | Tags: Decision-making, Human-agent, Individual performance, Induced stress, Time pressure
@inproceedings{soton485655,
title = {The effect of automated agents on individual performance under induced stress},
author = {Lokesh Singh and Gopal Ramchurn},
editor = {Jay Kalra},
url = {https://eprints.soton.ac.uk/485655/},
year = {2023},
date = {2023-11-01},
booktitle = {Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition)},
pages = {118–127},
publisher = {AHFE International},
abstract = {Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.},
keywords = {Decision-making, Human-agent, Individual performance, Induced stress, Time pressure},
pubstate = {published},
tppubtype = {inproceedings}
}
Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, vol. 49, no. 11, 2023, (Publisher Copyright: © 2023 The Author(s)).
Abstract | Links | BibTeX | Tags: Artificial Intelligence, machine learning, Oesophageal cancer multidisciplinary team
@article{soton479497b,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-11-01},
journal = {European Journal of Surgical Oncology},
volume = {49},
number = {11},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
© 2023 The Author(s)},
keywords = {Artificial Intelligence, machine learning, Oesophageal cancer multidisciplinary team},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub Response Miscellaneous
2023.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, chatgpt, european law, generative ai, Large Language Models, law and technology, technology policy
@misc{soton481740,
title = {AI large language models inquiry: TASHub Response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {Artificial Intelligence, chatgpt, european law, generative ai, Large Language Models, law and technology, technology policy},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub response Miscellaneous
2023.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, chatgpt, european law, generative ai, Large Language Models, law and technology, technology policy
@misc{soton481740b,
title = {AI large language models inquiry: TASHub response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {Artificial Intelligence, chatgpt, european law, generative ai, Large Language Models, law and technology, technology policy},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, 2023, (Publisher Copyright: copyright 2023 The Author(s)).
Abstract | Links | BibTeX | Tags: Artificial Intelligence, machine learning, Oesophageal cancer multidisciplinary team
@article{soton479497,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-07-01},
journal = {European Journal of Surgical Oncology},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
copyright 2023 The Author(s)},
keywords = {Artificial Intelligence, machine learning, Oesophageal cancer multidisciplinary team},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
Abioye, Ayodeji
University of Southampton, 2023.
Abstract | Links | BibTeX | Tags:
@phdthesis{soton479472,
title = {Multimodal speech and visual gesture control interface technique for small unmanned multirotor aircraft},
author = {Ayodeji Abioye},
url = {https://eprints.soton.ac.uk/479472/},
year = {2023},
date = {2023-07-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>
Abioye, Ayodeji; Naiseh, Mohammad; Hunt, William; Clark, Jediah R; Ramchurn, Sarvapali D; Soorati, Mohammad
The effect of data visualisation quality and task density on human-swarm interaction Proceedings Article
In: Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), IEEE, 2023.
Abstract | Links | BibTeX | Tags: featured_publication
@inproceedings{soton479970,
title = {The effect of data visualisation quality and task density on human-swarm interaction},
author = {Ayodeji Abioye and Mohammad Naiseh and William Hunt and Jediah R Clark and Sarvapali D Ramchurn and Mohammad Soorati},
url = {https://eprints.soton.ac.uk/479970/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
booktitle = {Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
publisher = {IEEE},
abstract = {Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.},
keywords = {featured_publication},
pubstate = {published},
tppubtype = {inproceedings}
}
Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.
Krook, Joshua; McAuley, Derek; Anderson, Stuart; Downer, John; Winter, Peter; Ramchurn, Sarvapali D
AI Foundation Models: initial review, CMA Consultation, TAS Hub Response Miscellaneous
2023.
Links | BibTeX | Tags: Artificial Intelligence, Competition policy, featured_publication, Foundation Models, Large Language Models, markets
@misc{soton477553,
title = {AI Foundation Models: initial review, CMA Consultation, TAS Hub Response},
author = {Joshua Krook and Derek McAuley and Stuart Anderson and John Downer and Peter Winter and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477553/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
publisher = {University of Southampton},
keywords = {Artificial Intelligence, Competition policy, featured_publication, Foundation Models, Large Language Models, markets},
pubstate = {published},
tppubtype = {misc}
}
Krook, Joshua; Downer, John; Winter, Peter; Williams, Jennifer; Ives, Jonathan; Bratu, Roxana; Sheir, Stephanie; Williams, Robin; Anderson, Stuart; Li, Phoebe; Ramamoorthy, Subramanian; Ramchurn, Sarvapali
AI regulation: a pro-innovation approach ? policy proposals: TASHub Response Miscellaneous
2023.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Consultation, innovation, Regulation, Trustworthy Autonomous Systems
@misc{soton478329,
title = {AI regulation: a pro-innovation approach ? policy proposals: TASHub Response},
author = {Joshua Krook and John Downer and Peter Winter and Jennifer Williams and Jonathan Ives and Roxana Bratu and Stephanie Sheir and Robin Williams and Stuart Anderson and Phoebe Li and Subramanian Ramamoorthy and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478329/},
year = {2023},
date = {2023-06-01},
publisher = {University of Southampton},
abstract = {Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence},
keywords = {Artificial Intelligence, Consultation, innovation, Regulation, Trustworthy Autonomous Systems},
pubstate = {published},
tppubtype = {misc}
}
Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence
Hunt, William; Ryan, Jack; Abioye, Ayodeji O; Ramchurn, Sarvapali D; Soorati, Mohammad D
Demonstrating performance benefits of human-swarm teaming Proceedings Article
In: Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems, pp. 3062–3064, International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS), 2023.
Abstract | Links | BibTeX | Tags: featured_publication
@inproceedings{soton479903,
title = {Demonstrating performance benefits of human-swarm teaming},
author = {William Hunt and Jack Ryan and Ayodeji O Abioye and Sarvapali D Ramchurn and Mohammad D Soorati},
url = {https://eprints.soton.ac.uk/479903/},
year = {2023},
date = {2023-05-01},
urldate = {2023-05-01},
booktitle = {Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems},
pages = {3062–3064},
publisher = {International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS)},
abstract = {Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.},
keywords = {featured_publication},
pubstate = {published},
tppubtype = {inproceedings}
}
Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.
Worrawichaipat, Phuriwat; Gerding, Enrico; Kaparias, Ioannis; Ramchurn, Sarvapali
Multi-agent signal-less intersection management with dynamic platoon formation Proceedings Article
In: 22nd International Conference on Autonomous Agents and Multiagent Systems (29/05/23 - 02/06/23), pp. 1542–1550, 2023.
Links | BibTeX | Tags: featured_publication
@inproceedings{soton478647,
title = {Multi-agent signal-less intersection management with dynamic platoon formation},
author = {Phuriwat Worrawichaipat and Enrico Gerding and Ioannis Kaparias and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478647/},
year = {2023},
date = {2023-05-01},
urldate = {2023-05-01},
booktitle = {22nd International Conference on Autonomous Agents and Multiagent Systems (29/05/23 - 02/06/23)},
pages = {1542–1550},
keywords = {featured_publication},
pubstate = {published},
tppubtype = {inproceedings}
}
Everett, Gregory; Beal, Ryan J; Matthews, Tim; Early, Joseph; Norman, Timothy J; Ramchurn, Sarvapali D
Inferring player location in sports matches: multi-agent spatial imputation from limited observations Miscellaneous
2023.
Abstract | Links | BibTeX | Tags: cs.LG, cs.MA
@misc{soton477020,
title = {Inferring player location in sports matches: multi-agent spatial imputation from limited observations},
author = {Gregory Everett and Ryan J Beal and Tim Matthews and Joseph Early and Timothy J Norman and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477020/},
year = {2023},
date = {2023-02-01},
abstract = {Understanding agent behaviour in Multi-Agent Systems (MAS) is an important problem in domains such as autonomous driving, disaster response, and sports analytics. Existing MAS problems typically use uniform timesteps with observations for all agents. In this work, we analyse the problem of agent location imputation, specifically posed in environments with non-uniform timesteps and limited agent observability (textttchar12695% missing values). Our approach uses Long Short-Term Memory and Graph Neural Network components to learn temporal and inter-agent patterns to predict the location of all agents at every timestep. We apply this to the domain of football (soccer) by imputing the location of all players in a game from sparse event data (e.g., shots and passes). Our model estimates player locations to within textttchar1266.9m; a textttchar12662% reduction in error from the best performing baseline. This approach facilitates downstream analysis tasks such as player physical metrics, player coverage, and team pitch control. Existing solutions to these tasks often require optical tracking data, which is expensive to obtain and only available to elite clubs. By imputing player locations from easy to obtain event data, we increase the accessibility of downstream tasks.},
keywords = {cs.LG, cs.MA},
pubstate = {published},
tppubtype = {misc}
}
Understanding agent behaviour in Multi-Agent Systems (MAS) is an important problem in domains such as autonomous driving, disaster response, and sports analytics. Existing MAS problems typically use uniform timesteps with observations for all agents. In this work, we analyse the problem of agent location imputation, specifically posed in environments with non-uniform timesteps and limited agent observability (textttchar12695% missing values). Our approach uses Long Short-Term Memory and Graph Neural Network components to learn temporal and inter-agent patterns to predict the location of all agents at every timestep. We apply this to the domain of football (soccer) by imputing the location of all players in a game from sparse event data (e.g., shots and passes). Our model estimates player locations to within textttchar1266.9m; a textttchar12662% reduction in error from the best performing baseline. This approach facilitates downstream analysis tasks such as player physical metrics, player coverage, and team pitch control. Existing solutions to these tasks often require optical tracking data, which is expensive to obtain and only available to elite clubs. By imputing player locations from easy to obtain event data, we increase the accessibility of downstream tasks.
2022
Ahmed, Sarah; Azim, Tayyaba; Early, Joseph Arthur; Ramchurn, Sarvapali
Revisiting deep fisher vectors: using fisher information to improve object classification Proceedings Article
In: British Machine Vision Conference (21/11/22 - 24/11/22), 2022.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton471260,
title = {Revisiting deep fisher vectors: using fisher information to improve object classification},
author = {Sarah Ahmed and Tayyaba Azim and Joseph Arthur Early and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/471260/},
year = {2022},
date = {2022-11-01},
booktitle = {British Machine Vision Conference (21/11/22 - 24/11/22)},
abstract = {Although deep learning models have become the gold standard in achieving outstanding results on a large variety of computer vision and machine learning tasks, the use of kernel methods has still not gone out of trend because of its potential to beat deep learning performances at a number of occasions. Given the potential of kernel techniques, prior works have also proposed the use of hybrid approaches combining deep learning with kernel learning to complement their respective strengths and weaknesses. This work develops this idea further by introducing an improved version of Fisher kernels derived from the deep Boltzmann machines (DBM). Our improved deep Fisher kernel (IDFK) utilises an approximation of the Fisher information matrix to derive improved Fisher vectors. We show IDFK can be utilised to retain a high degree of class separability, making it appropriate for classification and retrieval tasks. The efficacy of the proposed approach is evaluated on three benchmark data sets: MNIST, USPS and Alphanumeric, showing an improvement in classification performance over existing kernel approaches, and comparable performance to deep learning methods, but with much reduced computational costs. Using explainable AI methods, we also demonstrate why our IDFK leads to better classification performance.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Although deep learning models have become the gold standard in achieving outstanding results on a large variety of computer vision and machine learning tasks, the use of kernel methods has still not gone out of trend because of its potential to beat deep learning performances at a number of occasions. Given the potential of kernel techniques, prior works have also proposed the use of hybrid approaches combining deep learning with kernel learning to complement their respective strengths and weaknesses. This work develops this idea further by introducing an improved version of Fisher kernels derived from the deep Boltzmann machines (DBM). Our improved deep Fisher kernel (IDFK) utilises an approximation of the Fisher information matrix to derive improved Fisher vectors. We show IDFK can be utilised to retain a high degree of class separability, making it appropriate for classification and retrieval tasks. The efficacy of the proposed approach is evaluated on three benchmark data sets: MNIST, USPS and Alphanumeric, showing an improvement in classification performance over existing kernel approaches, and comparable performance to deep learning methods, but with much reduced computational costs. Using explainable AI methods, we also demonstrate why our IDFK leads to better classification performance.
Yazdanpanah, Vahid; Gerding, Enrico; Stein, Sebastian; Dastani, Mehdi; Jonker, Catholijn M; Norman, Timothy; Ramchurn, Sarvapali
Reasoning About Responsibility in Autonomous Systems: Challenges and Opportunities Journal Article
In: AI & Society, 2022.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Citizen-Centric AI Systems, human-agent collectives, Human-Centred AI, Multiagent Responsibility Reasoning, Multiagent Systems, Trustworthy Autonomous Systems
@article{soton471971,
title = {Reasoning About Responsibility in Autonomous Systems: Challenges and Opportunities},
author = {Vahid Yazdanpanah and Enrico Gerding and Sebastian Stein and Mehdi Dastani and Catholijn M Jonker and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/471971/},
year = {2022},
date = {2022-11-01},
journal = {AI & Society},
abstract = {Ensuring the trustworthiness of autonomous systems and artificial intelligenceensuremath<br/ensuremath>is an important interdisciplinary endeavour. In this position paper, we argue thatensuremath<br/ensuremath>this endeavour will benefit from technical advancements in capturing various forms of responsibility, and we present a comprehensive research agenda to achieve this. In particular, we argue that ensuring the reliability of autonomous system can take advantage of technical approaches for quantifying degrees of responsibility and for coordinating tasks based on that. Moreover, we deem that, in certifying the legality of an AI system, formal and computationally implementable notions of responsibility, blame, accountability, and liability are applicable for addressing potential responsibility gaps (i.e., situations in which a group is responsible, but individuals? responsibility may be unclear). This is a call to enable AI systems themselves, as well as those involved in the design, monitoring, and governance of AI systems, to represent and reason about who can be seen as responsible in prospect (e.g., for completing a task in future) and who can be seen as responsible retrospectively (e.g., for a failure that has already occurred). To that end, in this work, we show that across all stages of the design, development, and deployment of Trustworthy Autonomous Systems (TAS), responsibility reasoning should play a key role. This position paper is the first step towards establishing a road-map and research agenda on how the notion of responsibility can provide novel solution concepts for ensuring the reliability and legality of TAS and, as a result, enables an effective embedding of AI technologies into society.},
keywords = {Artificial Intelligence, Citizen-Centric AI Systems, human-agent collectives, Human-Centred AI, Multiagent Responsibility Reasoning, Multiagent Systems, Trustworthy Autonomous Systems},
pubstate = {published},
tppubtype = {article}
}
Ensuring the trustworthiness of autonomous systems and artificial intelligenceensuremath<br/ensuremath>is an important interdisciplinary endeavour. In this position paper, we argue thatensuremath<br/ensuremath>this endeavour will benefit from technical advancements in capturing various forms of responsibility, and we present a comprehensive research agenda to achieve this. In particular, we argue that ensuring the reliability of autonomous system can take advantage of technical approaches for quantifying degrees of responsibility and for coordinating tasks based on that. Moreover, we deem that, in certifying the legality of an AI system, formal and computationally implementable notions of responsibility, blame, accountability, and liability are applicable for addressing potential responsibility gaps (i.e., situations in which a group is responsible, but individuals? responsibility may be unclear). This is a call to enable AI systems themselves, as well as those involved in the design, monitoring, and governance of AI systems, to represent and reason about who can be seen as responsible in prospect (e.g., for completing a task in future) and who can be seen as responsible retrospectively (e.g., for a failure that has already occurred). To that end, in this work, we show that across all stages of the design, development, and deployment of Trustworthy Autonomous Systems (TAS), responsibility reasoning should play a key role. This position paper is the first step towards establishing a road-map and research agenda on how the notion of responsibility can provide novel solution concepts for ensuring the reliability and legality of TAS and, as a result, enables an effective embedding of AI technologies into society.
Early, Joseph; Deweese, Ying-Jung; Evers, Christine; Ramchurn, Sarvapali
Scene-to-Patch earth observation: multiple instance learning for land cover classification Miscellaneous
2022, (14 pages total (4 main content; 2 acknowledgments + citations; 8 appendices); 8 figures (2 main; 6 appendix); published at "Tackling Climate Change with Machine Learning: Workshop at NeurIPS 2022").
Abstract | Links | BibTeX | Tags: cs.CV, cs.LG
@misc{soton472853,
title = {Scene-to-Patch earth observation: multiple instance learning for land cover classification},
author = {Joseph Early and Ying-Jung Deweese and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/472853/},
year = {2022},
date = {2022-11-01},
abstract = {Land cover classification (LCC), and monitoring how land use changes over time, is an important process in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation data for LCC rely on fully-annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of LCC. In this study, we propose Scene-to-Patch models: an alternative LCC approach utilising Multiple Instance Learning (MIL) that requires only high-level scene labels. This enables much faster development of new datasets whilst still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using LCC for different scenarios. On the DeepGlobe-LCC dataset, our approach outperforms non-MIL baselines on both scene- and patch-level prediction. This work provides the foundation for expanding the use of LCC in climate change mitigation methods for technology, government, and academia.},
note = {14 pages total (4 main content; 2 acknowledgments + citations; 8 appendices); 8 figures (2 main; 6 appendix); published at "Tackling Climate Change with Machine Learning: Workshop at NeurIPS 2022"},
keywords = {cs.CV, cs.LG},
pubstate = {published},
tppubtype = {misc}
}
Land cover classification (LCC), and monitoring how land use changes over time, is an important process in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation data for LCC rely on fully-annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of LCC. In this study, we propose Scene-to-Patch models: an alternative LCC approach utilising Multiple Instance Learning (MIL) that requires only high-level scene labels. This enables much faster development of new datasets whilst still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using LCC for different scenarios. On the DeepGlobe-LCC dataset, our approach outperforms non-MIL baselines on both scene- and patch-level prediction. This work provides the foundation for expanding the use of LCC in climate change mitigation methods for technology, government, and academia.
Parnell, Katie; Fischer, Joel E; Clark, Jediah R; Bodenmann, Adrian; Trigo, Maria Jose Galvez; Brito, Mario; Soorati, Mohammad Divband; Plant, Katherine; Ramchurn, Sarvapali
Trustworthy UAV relationships: Applying the Schema Action World taxonomy to UAVs and UAV swarm operations Journal Article
In: International Journal of Human-Computer Interaction, 2022.
Abstract | Links | BibTeX | Tags:
@article{soton468839,
title = {Trustworthy UAV relationships: Applying the Schema Action World taxonomy to UAVs and UAV swarm operations},
author = {Katie Parnell and Joel E Fischer and Jediah R Clark and Adrian Bodenmann and Maria Jose Galvez Trigo and Mario Brito and Mohammad Divband Soorati and Katherine Plant and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/468839/},
year = {2022},
date = {2022-07-01},
journal = {International Journal of Human-Computer Interaction},
abstract = {Human Factors play a significant role inthe development and integration of avionic systems to ensure that they are trusted and can be used effectively. As Unoccupied Aerial Vehicle (UAV) technology becomes increasingly important to the aviation domain this holds true. This study aims to gain an understanding of UAV operators?trust requirements when piloting UAVs by utilising a popular aviation interview methodology (Schema World Action Research Method), in combination with key questions on trust identified from the literature. Interviews were conducted with six UAVoperators, with a range of experience. This identified the importance of past experience to trust and the expectations that operators hold. Recommendations are made that target training to inform experience, in addition to the equipment, procedures and organisational standards that can aid in developing trustworthy systems. The methodology that was developed shows promise for capturing trust within human-automation interactions},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Human Factors play a significant role inthe development and integration of avionic systems to ensure that they are trusted and can be used effectively. As Unoccupied Aerial Vehicle (UAV) technology becomes increasingly important to the aviation domain this holds true. This study aims to gain an understanding of UAV operators?trust requirements when piloting UAVs by utilising a popular aviation interview methodology (Schema World Action Research Method), in combination with key questions on trust identified from the literature. Interviews were conducted with six UAVoperators, with a range of experience. This identified the importance of past experience to trust and the expectations that operators hold. Recommendations are made that target training to inform experience, in addition to the equipment, procedures and organisational standards that can aid in developing trustworthy systems. The methodology that was developed shows promise for capturing trust within human-automation interactions
Soorati, Mohammad Divband; Gerding, Enrico; Marchioni, Enrico; Naumov, Pavel; Norman, Timothy; Ramchurn, Sarvapali; Rastegari, Baharak; Sobey, Adam; Stein, Sebastian; Tarapore, Danesh; Yazdanpanah, Vahid; Zhang, Jie
From Intelligent Agents to Trustworthy Human-Centred Multiagent Systems Journal Article
In: AI Communications, 2022.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Distributed Artificial Intelligence, Intelligent Agents, Multiagent Systems, Trustworthy Autonomous Systems
@article{soton467975,
title = {From Intelligent Agents to Trustworthy Human-Centred Multiagent Systems},
author = {Mohammad Divband Soorati and Enrico Gerding and Enrico Marchioni and Pavel Naumov and Timothy Norman and Sarvapali Ramchurn and Baharak Rastegari and Adam Sobey and Sebastian Stein and Danesh Tarapore and Vahid Yazdanpanah and Jie Zhang},
url = {https://eprints.soton.ac.uk/467975/},
year = {2022},
date = {2022-07-01},
journal = {AI Communications},
abstract = {The Agents, Interaction and Complexity research group at the University of Southampton has a long track record of research in multiagent systems (MAS). We have made substantial scientific contributions across learning in MAS, game-theoretic techniques for coordinating agent systems, and formal methods for representation and reasoning. We highlight key results achieved by the group and elaborate on recent work and open research challenges in developing trustworthy autonomous systems and deploying human-centred AI systems that aim to support societal good.},
keywords = {Artificial Intelligence, Distributed Artificial Intelligence, Intelligent Agents, Multiagent Systems, Trustworthy Autonomous Systems},
pubstate = {published},
tppubtype = {article}
}
The Agents, Interaction and Complexity research group at the University of Southampton has a long track record of research in multiagent systems (MAS). We have made substantial scientific contributions across learning in MAS, game-theoretic techniques for coordinating agent systems, and formal methods for representation and reasoning. We highlight key results achieved by the group and elaborate on recent work and open research challenges in developing trustworthy autonomous systems and deploying human-centred AI systems that aim to support societal good.
Bossens, David; Ramchurn, Sarvapali; Tarapore, Danesh
Resilient robot teams: a review integrating decentralised control, change-detection, and learning Miscellaneous
2022.
Abstract | Links | BibTeX | Tags:
@misc{soton457101,
title = {Resilient robot teams: a review integrating decentralised control, change-detection, and learning},
author = {David Bossens and Sarvapali Ramchurn and Danesh Tarapore},
url = {https://eprints.soton.ac.uk/457101/},
year = {2022},
date = {2022-06-01},
journal = {Current Robotics Reports},
abstract = {Purpose of review: This paper reviews opportunities and challenges for decentralised control, change-detection, and learning in the context of resilient robot teams.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Recent findings: Exogenous fault detection methods can provide a generic detection or a specific diagnosis with a recovery solution. Robot teams can perform active and distributed sensing for detecting changes in the environment, including identifying and tracking dynamic anomalies, as well as collaboratively mapping dynamic environments. Resilient methods for decentralised control have been developed in learning perception-action-communication loops, multi-agent reinforcement learning, embodied evolution, offline evolution with online adaptation, explicit task allocation, and stigmergy in swarm robotics.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Summary: Remaining challenges for resilient robot teams are integrating change-detection and trial-and-error learning methods, obtaining reliable performance evaluations under constrained evaluation time, improving the safety of resilient robot teams, theoretical results demonstrating rapid adaptation to given environmental perturbations, and designing realistic and compelling case studies.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Purpose of review: This paper reviews opportunities and challenges for decentralised control, change-detection, and learning in the context of resilient robot teams.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Recent findings: Exogenous fault detection methods can provide a generic detection or a specific diagnosis with a recovery solution. Robot teams can perform active and distributed sensing for detecting changes in the environment, including identifying and tracking dynamic anomalies, as well as collaboratively mapping dynamic environments. Resilient methods for decentralised control have been developed in learning perception-action-communication loops, multi-agent reinforcement learning, embodied evolution, offline evolution with online adaptation, explicit task allocation, and stigmergy in swarm robotics.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Summary: Remaining challenges for resilient robot teams are integrating change-detection and trial-and-error learning methods, obtaining reliable performance evaluations under constrained evaluation time, improving the safety of resilient robot teams, theoretical results demonstrating rapid adaptation to given environmental perturbations, and designing realistic and compelling case studies.
Early, Joseph; Bewley, Tom; Evers, Christine; Ramchurn, Sarvapali
Non-markovian reward modelling from trajectory labels via interpretable multiple instance learning Journal Article
In: arXiv, 2022, (20 pages (9 main content; 2 references; 9 appendix). 11 figures (8 main content; 3 appendix)).
Abstract | Links | BibTeX | Tags: cs.AI, cs.LG
@article{soton458023,
title = {Non-markovian reward modelling from trajectory labels via interpretable multiple instance learning},
author = {Joseph Early and Tom Bewley and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/458023/},
year = {2022},
date = {2022-05-01},
journal = {arXiv},
abstract = {We generalise the problem of reward modelling (RM) for reinforcement learning (RL) to handle non-Markovian rewards. Existing work assumes that human evaluators observe each step in a trajectory independently when providing feedback on agent behaviour. In this work, we remove this assumption, extending RM to include hidden state information that captures temporal dependencies in human assessment of trajectories. We then show how RM can be approached as a multiple instance learning (MIL) problem, and develop new MIL models that are able to capture the time dependencies in labelled trajectories. We demonstrate on a range of RL tasks that our novel MIL models can reconstruct reward functions to a high level of accuracy, and that they provide interpretable learnt hidden information that can be used to train high-performing agent policies.},
note = {20 pages (9 main content; 2 references; 9 appendix). 11 figures (8 main content; 3 appendix)},
keywords = {cs.AI, cs.LG},
pubstate = {published},
tppubtype = {article}
}
We generalise the problem of reward modelling (RM) for reinforcement learning (RL) to handle non-Markovian rewards. Existing work assumes that human evaluators observe each step in a trajectory independently when providing feedback on agent behaviour. In this work, we remove this assumption, extending RM to include hidden state information that captures temporal dependencies in human assessment of trajectories. We then show how RM can be approached as a multiple instance learning (MIL) problem, and develop new MIL models that are able to capture the time dependencies in labelled trajectories. We demonstrate on a range of RL tasks that our novel MIL models can reconstruct reward functions to a high level of accuracy, and that they provide interpretable learnt hidden information that can be used to train high-performing agent policies.
Buermann, Jan; Georgiev, Dimitar; Gerding, Enrico; Hill, Lewis; Malik, Obaid; Pop, Alexandru; Pun, Matthew; Ramchurn, Sarvapali; Salisbury, Elliot; Stojanovic, Ivan
An agent-based simulator for maritime transport decarbonisation: Demonstration track Proceedings Article
In: 21st International Conference on Autonomous Agents and Multiagent Systems (09/05/22 - 13/05/22), pp. 1890–1892, 2022.
Abstract | Links | BibTeX | Tags: Agent-Based Modelling and Simulation: Applications & Analysis, Emergent Behaviour, Integration of Agent-Based and Other Technologies, Simulation Techniques, Tools and Platforms
@inproceedings{soton456716,
title = {An agent-based simulator for maritime transport decarbonisation: Demonstration track},
author = {Jan Buermann and Dimitar Georgiev and Enrico Gerding and Lewis Hill and Obaid Malik and Alexandru Pop and Matthew Pun and Sarvapali Ramchurn and Elliot Salisbury and Ivan Stojanovic},
url = {https://eprints.soton.ac.uk/456716/},
year = {2022},
date = {2022-05-01},
booktitle = {21st International Conference on Autonomous Agents and Multiagent Systems (09/05/22 - 13/05/22)},
pages = {1890–1892},
abstract = {Greenhouse gas (GHG) emission reduction is an important and necessary goal; currently, different policies to reduce GHG emissions in maritime transport are being discussed. Amongst policies, like carbon taxes or carbon intensity targets, it is hard to determine which policies can successfully reduce GHG emissions while allowing the industry to be profitable. We introduce an agent-based maritime transport simulator to investigate the effectiveness of two decarbonisation policies by simulating a maritime transport operator?s trade pattern and fleet make-up changes as a reaction to taxation and fixed targets. This first of its kind simulator allows to compare and quantify the difference of carbon reduction policies and how they affect shipping operations.},
keywords = {Agent-Based Modelling and Simulation: Applications & Analysis, Emergent Behaviour, Integration of Agent-Based and Other Technologies, Simulation Techniques, Tools and Platforms},
pubstate = {published},
tppubtype = {inproceedings}
}
Greenhouse gas (GHG) emission reduction is an important and necessary goal; currently, different policies to reduce GHG emissions in maritime transport are being discussed. Amongst policies, like carbon taxes or carbon intensity targets, it is hard to determine which policies can successfully reduce GHG emissions while allowing the industry to be profitable. We introduce an agent-based maritime transport simulator to investigate the effectiveness of two decarbonisation policies by simulating a maritime transport operator?s trade pattern and fleet make-up changes as a reaction to taxation and fixed targets. This first of its kind simulator allows to compare and quantify the difference of carbon reduction policies and how they affect shipping operations.
Rigas, Emmanouil S; Gerding, Enrico H; Stein, Sebastian; Ramchurn, Sarvapali D; Bassiliades, Nick
Mechanism design for efficient offline and online allocation of electric vehicles to charging stations Journal Article
In: Energies, vol. 15, no. 5, 2022, (Funding Information: Funding: This research study was co-financed by Greece and the European Union (European Social Fund–ESF) through the Operational Programme ?Human Resources Development, Education and Lifelong Learning? in the context of the project ?Reinforcement of Postdoctoral Researchers-2nd Cycle? (MIS-5033021), implemented by State Scholarships Foundation (IKY). Copyright 2022 Elsevier B.V., All rights reserved.).
Abstract | Links | BibTeX | Tags: Charging, Electric Vehicles, Fixed price, mechanism design, Scheduling, VCG
@article{soton455806,
title = {Mechanism design for efficient offline and online allocation of electric vehicles to charging stations},
author = {Emmanouil S Rigas and Enrico H Gerding and Sebastian Stein and Sarvapali D Ramchurn and Nick Bassiliades},
url = {https://eprints.soton.ac.uk/455806/},
year = {2022},
date = {2022-03-01},
journal = {Energies},
volume = {15},
number = {5},
abstract = {ensuremath<pensuremath>The industry related to electric vehicles (EVs) has seen a substantial increase in recent years, as such vehicles have the ability to significantly reduce total COensuremath<subensuremath>2ensuremath</subensuremath> emissions and the related global warming effect. In this paper, we focus on the problem of allocating EVs to charging stations, scheduling and pricing their charging. Specifically, we developed a Mixed Integer Program (MIP) which executes offline and optimally allocates EVs to charging stations. On top, we propose two alternative mechanisms to price the electricity the EVs charge. The first mechanism is a typical fixed-price one, while the second is a variation of the Vickrey?Clark?Groves (VCG) mechanism. We also developed online solutions that incrementally call the MIP-based algorithm and solve it for branches of EVs. In all cases, the EVs? aim is to minimize the price to pay and the impact on their driving schedule, acting as self-interested agents. We conducted a thorough empirical evaluation of our mechanisms and we observed that they had satisfactory scalability. Additionally, the VCG mechanism achieved an up to 2.2% improvement in terms of the number of vehicles that were charged compared to the fixed-price one and, in cases where the stations were congested, it calculated higher prices for the EVs and provided a higher profit for the stations, but lower utility to the EVs. However, in a theoretical evaluation, we proved that the variant of the VCG mechanism being proposed in this paper still guaranteed truthful reporting of the EVs? preferences. In contrast, the fixed-price one was found to be vulnerable to agents? strategic behavior as non-truthful EVs can charge instead of truthful ones. Finally, we observed the online algorithms to be, on average, at 95.6% of the offline ones in terms of the average number of serviced EVs.ensuremath</pensuremath>},
note = {Funding Information:
Funding: This research study was co-financed by Greece and the European Union (European Social Fund–ESF) through the Operational Programme ?Human Resources Development, Education and Lifelong Learning? in the context of the project ?Reinforcement of Postdoctoral Researchers-2nd Cycle? (MIS-5033021), implemented by State Scholarships Foundation (IKY).
Copyright 2022 Elsevier B.V., All rights reserved.},
keywords = {Charging, Electric Vehicles, Fixed price, mechanism design, Scheduling, VCG},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>The industry related to electric vehicles (EVs) has seen a substantial increase in recent years, as such vehicles have the ability to significantly reduce total COensuremath<subensuremath>2ensuremath</subensuremath> emissions and the related global warming effect. In this paper, we focus on the problem of allocating EVs to charging stations, scheduling and pricing their charging. Specifically, we developed a Mixed Integer Program (MIP) which executes offline and optimally allocates EVs to charging stations. On top, we propose two alternative mechanisms to price the electricity the EVs charge. The first mechanism is a typical fixed-price one, while the second is a variation of the Vickrey?Clark?Groves (VCG) mechanism. We also developed online solutions that incrementally call the MIP-based algorithm and solve it for branches of EVs. In all cases, the EVs? aim is to minimize the price to pay and the impact on their driving schedule, acting as self-interested agents. We conducted a thorough empirical evaluation of our mechanisms and we observed that they had satisfactory scalability. Additionally, the VCG mechanism achieved an up to 2.2% improvement in terms of the number of vehicles that were charged compared to the fixed-price one and, in cases where the stations were congested, it calculated higher prices for the EVs and provided a higher profit for the stations, but lower utility to the EVs. However, in a theoretical evaluation, we proved that the variant of the VCG mechanism being proposed in this paper still guaranteed truthful reporting of the EVs? preferences. In contrast, the fixed-price one was found to be vulnerable to agents? strategic behavior as non-truthful EVs can charge instead of truthful ones. Finally, we observed the online algorithms to be, on average, at 95.6% of the offline ones in terms of the average number of serviced EVs.ensuremath</pensuremath>
Early, Joseph; Evers, Christine; Ramchurn, Sarvapali
Model agnostic interpretability for multiple instance learning Proceedings Article
In: International Conference on Learning Representations 2022 (25/04/22 - 29/04/22), 2022, (25 pages (9 content, 2 acknowledgement + references, 14 appendix). 16 figures (3 main content, 13 appendix). Submitted and accepted to ICLR 22, see http://openreview.net/forum?id=KSSfF5lMIAg . Revision: added additional acknowledgements).
Abstract | Links | BibTeX | Tags: cs.AI, cs.LG
@inproceedings{soton454952,
title = {Model agnostic interpretability for multiple instance learning},
author = {Joseph Early and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/454952/},
year = {2022},
date = {2022-01-01},
booktitle = {International Conference on Learning Representations 2022 (25/04/22 - 29/04/22)},
abstract = {In Multiple Instance Learning (MIL), models are trained using bags of instances, where only a single label is provided for each bag. A bag label is often only determined by a handful of key instances within a bag, making it difficult to interpret what information a classifier is using to make decisions. In this work, we establish the key requirements for interpreting MIL models. We then go on to develop several model-agnostic approaches that meet these requirements. Our methods are compared against existing inherently interpretable MIL models on several datasets, and achieve an increase in interpretability accuracy of up to 30%. We also examine the ability of the methods to identify interactions between instances and scale to larger datasets, improving their applicability to real-world problems.},
note = {25 pages (9 content, 2 acknowledgement + references, 14 appendix). 16 figures (3 main content, 13 appendix). Submitted and accepted to ICLR 22, see http://openreview.net/forum?id=KSSfF5lMIAg . Revision: added additional acknowledgements},
keywords = {cs.AI, cs.LG},
pubstate = {published},
tppubtype = {inproceedings}
}
In Multiple Instance Learning (MIL), models are trained using bags of instances, where only a single label is provided for each bag. A bag label is often only determined by a handful of key instances within a bag, making it difficult to interpret what information a classifier is using to make decisions. In this work, we establish the key requirements for interpreting MIL models. We then go on to develop several model-agnostic approaches that meet these requirements. Our methods are compared against existing inherently interpretable MIL models on several datasets, and achieve an increase in interpretability accuracy of up to 30%. We also examine the ability of the methods to identify interactions between instances and scale to larger datasets, improving their applicability to real-world problems.
2021
Ramchurn, Sarvapali; Middleton, Stuart; McAuley, Derek; Webb, Helena; Hyde, Richard; Lisinska, Justyna
A Response to Draft Online Safety Bill: A call for evidence from the Joint Committee Technical Report
no. 10.18742/pub01-060, 2021.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Online Harms
@techreport{soton451428,
title = {A Response to Draft Online Safety Bill: A call for evidence from the Joint Committee},
author = {Sarvapali Ramchurn and Stuart Middleton and Derek McAuley and Helena Webb and Richard Hyde and Justyna Lisinska},
url = {https://eprints.soton.ac.uk/451428/},
year = {2021},
date = {2021-09-01},
number = {10.18742/pub01-060},
abstract = {This report is the Trustworthy Autonomous Hub (TAS-hub) response to the call for evidence from the Joint Committee on the Draft Online Safety Bill. The Joint Committee was established to consider the Government's draft Bill to establish a new regulatory framework to tackle harmful content online.},
keywords = {Artificial Intelligence, Online Harms},
pubstate = {published},
tppubtype = {techreport}
}
This report is the Trustworthy Autonomous Hub (TAS-hub) response to the call for evidence from the Joint Committee on the Draft Online Safety Bill. The Joint Committee was established to consider the Government's draft Bill to establish a new regulatory framework to tackle harmful content online.
Ramchurn, Sarvapali; Stein, Sebastian; Jennings, Nicholas R
Trustworthy human-AI partnerships Journal Article
In: iScience, vol. 24, no. 8, 2021.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Human-Computer Interaction, Sociology
@article{soton450597,
title = {Trustworthy human-AI partnerships},
author = {Sarvapali Ramchurn and Sebastian Stein and Nicholas R Jennings},
url = {https://eprints.soton.ac.uk/450597/},
year = {2021},
date = {2021-08-01},
journal = {iScience},
volume = {24},
number = {8},
abstract = {In this paper, we foreground some of the key research challenges that arise in the design of trustworthy human-AI partnerships. In particular, we focus on the challenges in designing human-AI partnerships that need to be addressed to help humans and organisations trust their machine counterparts individually or as a collective (e.g., as robot teams or groups of software agents). We also aim to identify the risks associated with human-AI partnerships and therefore determine the associated measures to mitigate these risks. By so doing, we will trigger new avenues of research that will address the key barriers to the adoption of AI-based systems more widely in our daily lives and in industry.},
keywords = {Artificial Intelligence, Human-Computer Interaction, Sociology},
pubstate = {published},
tppubtype = {article}
}
In this paper, we foreground some of the key research challenges that arise in the design of trustworthy human-AI partnerships. In particular, we focus on the challenges in designing human-AI partnerships that need to be addressed to help humans and organisations trust their machine counterparts individually or as a collective (e.g., as robot teams or groups of software agents). We also aim to identify the risks associated with human-AI partnerships and therefore determine the associated measures to mitigate these risks. By so doing, we will trigger new avenues of research that will address the key barriers to the adoption of AI-based systems more widely in our daily lives and in industry.
Ramchurn, Sarvapali; Mousavi, Mohammad Reza; Toliyat, Seyed Mohammad Hossein; Kleinman, Mark; Lisinska, Justyna; Sempreboni, Diego; Stein, Sebastian; Gerding, Enrico; Gomer, Richard; DÁmore, Francesco
The future of connected and automated mobility in the UK: call for evidence Technical Report
no. 10.5258/SOTON/P0097, 2021, (The UKRI TAS Hub assembles a team from the Universities of Southampton, Nottingham and King?s College London. The Hub sits at the centre of the pounds33M Trustworthy Autonomous Systems Programme, funded by the UKRI Strategic Priorities Fund. The role of the TAS Hub is to coordinate and work with six research nodes to establish a collaborative platform for the UK to enable the development of socially beneficial autonomous systems that are both trustworthy in principle and trusted in practice by individuals, society and government. Read more about the TAS Hub at https://www.tas.ac.uk/aboutus/overview/).
Abstract | Links | BibTeX | Tags:
@techreport{soton450228,
title = {The future of connected and automated mobility in the UK: call for evidence},
author = {Sarvapali Ramchurn and Mohammad Reza Mousavi and Seyed Mohammad Hossein Toliyat and Mark Kleinman and Justyna Lisinska and Diego Sempreboni and Sebastian Stein and Enrico Gerding and Richard Gomer and Francesco DÁmore},
editor = {Wassim Dbouk},
url = {https://eprints.soton.ac.uk/450228/},
year = {2021},
date = {2021-07-01},
number = {10.5258/SOTON/P0097},
publisher = {University of Southampton},
abstract = {This report is a response to the call for evidence from the Department for Business, Energy & Industrial Strategy and the Centre for Connected and Autonomous Vehicles on the future of connected and automated mobility in the UK.ensuremath<br/ensuremath>Executive Summary:Despite relative weaknesses in global collaboration and co-creation platforms, smart road and communication infrastructure, urban planning, and public awareness, the United Kingdom (UK) has a substantial strength in the area of Connected and Automated Mobility (CAM) by investing in research and innovation platforms for developing the underlying technologies, creating impact, and co-creation leading to innovative solutions. Many UK legal and policymaking initiatives in this domain are world leading. To sustain the UK?s leading position, we make the following recommendations:? The development of financial and policy-related incentive schemes for research and innovation in the foundations and applications of autonomous systems as well as schemes for proof of concepts, and commercialisation.? Supporting policy and standardisation initiatives as well as engagement and community-building activities to increase public awareness and trust.? Giving greater attention to integrating CAM/Connected Autonomous Shared Electric vehicles (CASE) policy with related government priorities for mobility, including supporting active transport and public transport, and improving air quality.? Further investment in updating liability and risk models and coming up with innovative liability schemes covering the Autonomous Vehicles (AVs) ecosystem.? Investing in training and retraining of the work force in the automotive, mobility, and transport sectors, particularly with skills concerningArtificial Intelligence (AI), software and computer systems, in order to ensure employability and an adequate response to the drastically changing industrial landscape},
note = {The UKRI TAS Hub assembles a team from the Universities of Southampton, Nottingham and King?s College London. The Hub sits at the centre of the pounds33M Trustworthy Autonomous Systems Programme, funded by the UKRI Strategic Priorities Fund.
The role of the TAS Hub is to coordinate and work with six research nodes to establish a collaborative platform for the UK to enable the development of socially beneficial autonomous systems that are both trustworthy in principle and trusted in practice by individuals, society and government. Read more about the TAS Hub at https://www.tas.ac.uk/aboutus/overview/},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
This report is a response to the call for evidence from the Department for Business, Energy & Industrial Strategy and the Centre for Connected and Autonomous Vehicles on the future of connected and automated mobility in the UK.ensuremath<br/ensuremath>Executive Summary:Despite relative weaknesses in global collaboration and co-creation platforms, smart road and communication infrastructure, urban planning, and public awareness, the United Kingdom (UK) has a substantial strength in the area of Connected and Automated Mobility (CAM) by investing in research and innovation platforms for developing the underlying technologies, creating impact, and co-creation leading to innovative solutions. Many UK legal and policymaking initiatives in this domain are world leading. To sustain the UK?s leading position, we make the following recommendations:? The development of financial and policy-related incentive schemes for research and innovation in the foundations and applications of autonomous systems as well as schemes for proof of concepts, and commercialisation.? Supporting policy and standardisation initiatives as well as engagement and community-building activities to increase public awareness and trust.? Giving greater attention to integrating CAM/Connected Autonomous Shared Electric vehicles (CASE) policy with related government priorities for mobility, including supporting active transport and public transport, and improving air quality.? Further investment in updating liability and risk models and coming up with innovative liability schemes covering the Autonomous Vehicles (AVs) ecosystem.? Investing in training and retraining of the work force in the automotive, mobility, and transport sectors, particularly with skills concerningArtificial Intelligence (AI), software and computer systems, in order to ensure employability and an adequate response to the drastically changing industrial landscape
Worrawichaipat, Phuriwat; Gerding, Enrico; Kaparias, Ioannis; Ramchurn, Sarvapali
Resilient intersection management with multi-vehicle collision avoidance Journal Article
In: Frontiers in Sustainable Cities, vol. 3, 2021.
Abstract | Links | BibTeX | Tags: Computer science, intersection management, multi-agent systems, simulation experiments, Transportation
@article{soton449675,
title = {Resilient intersection management with multi-vehicle collision avoidance},
author = {Phuriwat Worrawichaipat and Enrico Gerding and Ioannis Kaparias and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449675/},
year = {2021},
date = {2021-06-01},
journal = {Frontiers in Sustainable Cities},
volume = {3},
abstract = {In this paper, we propose a novel decentralised agent-based mechanism for road intersection management for connected autonomous vehicles. In our work we focus on road obstructions causing major traffic delays. In doing so, we propose the first decentralised mechanism able to maximise the overall vehicle throughput at intersections in the presence of obstructions. The distributed algorithm transfers most of the computational cost from the intersection manager to the driving agents, thereby improving scalability. Our realistic empirical experiments using SUMO show that, when an obstacle is located at the entrance or in the middle of the intersection, existing state of the art algorithms and traffic lights show a reduced throughput of 65?90% from the optimal point without obstructions while our mechanism can maintain the throughput upensuremath<br/ensuremath>Q7 to 94?99%.},
keywords = {Computer science, intersection management, multi-agent systems, simulation experiments, Transportation},
pubstate = {published},
tppubtype = {article}
}
In this paper, we propose a novel decentralised agent-based mechanism for road intersection management for connected autonomous vehicles. In our work we focus on road obstructions causing major traffic delays. In doing so, we propose the first decentralised mechanism able to maximise the overall vehicle throughput at intersections in the presence of obstructions. The distributed algorithm transfers most of the computational cost from the intersection manager to the driving agents, thereby improving scalability. Our realistic empirical experiments using SUMO show that, when an obstacle is located at the entrance or in the middle of the intersection, existing state of the art algorithms and traffic lights show a reduced throughput of 65?90% from the optimal point without obstructions while our mechanism can maintain the throughput upensuremath<br/ensuremath>Q7 to 94?99%.
Capezzuto, Luca; Tarapore, Danesh; Ramchurn, Sarvapali
Large-scale, dynamic and distributed coalition formation with spatial and†temporal constraints Journal Article
In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), pp. 108–125, 2021.
Abstract | Links | BibTeX | Tags:
@article{soton452050,
title = {Large-scale, dynamic and distributed coalition formation with spatial and†temporal constraints},
author = {Luca Capezzuto and Danesh Tarapore and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/452050/},
year = {2021},
date = {2021-05-01},
journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)},
pages = {108–125},
abstract = {The†Coalition Formation with Spatial and Temporal constraints Problem†(CFSTP) is a multi-agent task allocation problem in which few agents have to perform many tasks, each with its deadline and workload. To maximize the number of completed tasks, the agents need to cooperate by forming, disbanding and reforming coalitions. The original mathematical programming formulation of the CFSTP is difficult to implement, since it is lengthy and based on the problematic Big-M method. In this paper, we propose a compact and easy-to-implement formulation. Moreover, we design D-CTS, a distributed version of the state-of-the-art CFSTP algorithm. Using public London Fire Brigade records, we create a dataset with 347588 tasks and a test framework that simulates the mobilization of firefighters in dynamic environments. In problems with up†to 150 agents and 3000 tasks, compared to DSA-SDP, a state-of-the-art distributed algorithm, D-CTS completes†3.79%$pm$[42.22%,1.96%]†more tasks, and is one order of magnitude more efficient in terms of communication overhead and time complexity. D-CTS sets the first large-scale, dynamic and distributed CFSTP benchmark.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The†Coalition Formation with Spatial and Temporal constraints Problem†(CFSTP) is a multi-agent task allocation problem in which few agents have to perform many tasks, each with its deadline and workload. To maximize the number of completed tasks, the agents need to cooperate by forming, disbanding and reforming coalitions. The original mathematical programming formulation of the CFSTP is difficult to implement, since it is lengthy and based on the problematic Big-M method. In this paper, we propose a compact and easy-to-implement formulation. Moreover, we design D-CTS, a distributed version of the state-of-the-art CFSTP algorithm. Using public London Fire Brigade records, we create a dataset with 347588 tasks and a test framework that simulates the mobilization of firefighters in dynamic environments. In problems with up†to 150 agents and 3000 tasks, compared to DSA-SDP, a state-of-the-art distributed algorithm, D-CTS completes†3.79%$pm$[42.22%,1.96%]†more tasks, and is one order of magnitude more efficient in terms of communication overhead and time complexity. D-CTS sets the first large-scale, dynamic and distributed CFSTP benchmark.ensuremath<br/ensuremath>
Capezzuto, Luca; Tarapore, Danesh; Ramchurn, Sarvapali D
Anytime and efficient multi-agent coordination for disaster response Journal Article
In: SN Computer Science, vol. 2, no. 3, 2021.
Abstract | Links | BibTeX | Tags:
@article{soton467373,
title = {Anytime and efficient multi-agent coordination for disaster response},
author = {Luca Capezzuto and Danesh Tarapore and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/467373/},
year = {2021},
date = {2021-03-01},
journal = {SN Computer Science},
volume = {2},
number = {3},
abstract = {The Coalition Formation with Spatial and Temporal constraints Problem (CFSTP) is a multi-agent task allocation problem where the tasks are spatially distributed, with deadlines and workloads, and the number of agents is typically much smaller than the number of tasks. To maximise the number of completed tasks, the agents may have to schedule coalitions. The state-of-the-art CFSTP solver, the Coalition Formation with Look-Ahead (CFLA) algorithm, has two main limitations. First, its time complexity is exponential with the number of agents. Second, as we show, its look-ahead technique is not effective in real-world scenarios, such as open multi-agent systems, where new tasks can appear at any time. In this work, we study its design and define a variant, called Coalition Formation with Improved Look-Ahead (CFLA2), which achieves better performance. Since we cannot eliminate the limitations of CFLA in CFLA2, we also develop a novel algorithm to solve the CFSTP, the first to be simultaneously anytime, efficient and with convergence guarantee, called Cluster-based Task Scheduling (CTS). In tests where the look-ahead technique is highly effective, CTS completes up to 30% (resp. 10%) more tasks than CFLA (resp. CFLA2) while being up to 4 orders of magnitude faster. We also propose S-CTS, a simplified but parallel variant of CTS with even lower time complexity. Using scenarios generated by the RoboCup Rescue Simulation, we show that S-CTS is at most 10% less performing than high-performance algorithms such as Binary Max-Sum and DSA, but up to 2 orders of magnitude faster. Our results affirm CTS as the new state-of-the-art algorithm to solve the CFSTP.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The Coalition Formation with Spatial and Temporal constraints Problem (CFSTP) is a multi-agent task allocation problem where the tasks are spatially distributed, with deadlines and workloads, and the number of agents is typically much smaller than the number of tasks. To maximise the number of completed tasks, the agents may have to schedule coalitions. The state-of-the-art CFSTP solver, the Coalition Formation with Look-Ahead (CFLA) algorithm, has two main limitations. First, its time complexity is exponential with the number of agents. Second, as we show, its look-ahead technique is not effective in real-world scenarios, such as open multi-agent systems, where new tasks can appear at any time. In this work, we study its design and define a variant, called Coalition Formation with Improved Look-Ahead (CFLA2), which achieves better performance. Since we cannot eliminate the limitations of CFLA in CFLA2, we also develop a novel algorithm to solve the CFSTP, the first to be simultaneously anytime, efficient and with convergence guarantee, called Cluster-based Task Scheduling (CTS). In tests where the look-ahead technique is highly effective, CTS completes up to 30% (resp. 10%) more tasks than CFLA (resp. CFLA2) while being up to 4 orders of magnitude faster. We also propose S-CTS, a simplified but parallel variant of CTS with even lower time complexity. Using scenarios generated by the RoboCup Rescue Simulation, we show that S-CTS is at most 10% less performing than high-performance algorithms such as Binary Max-Sum and DSA, but up to 2 orders of magnitude faster. Our results affirm CTS as the new state-of-the-art algorithm to solve the CFSTP.
Ortega, Andre P; Ramchurn, Sarvapali; Tran-Thanh, Long; Merrett, Geoff
Partner selection in self-organised wireless sensor networks for opportunistic energy negotiation: A multi-armed bandit based approach Journal Article
In: Ad Hoc Networks, vol. 112, 2021.
Abstract | Links | BibTeX | Tags: Agent-based sensor network, Automated negotiation, Energy management, Multi-armed bandit based learning, Reinforcement Learning, Wireless sensor networks
@article{soton445733,
title = {Partner selection in self-organised wireless sensor networks for opportunistic energy negotiation: A multi-armed bandit based approach},
author = {Andre P Ortega and Sarvapali Ramchurn and Long Tran-Thanh and Geoff Merrett},
url = {https://eprints.soton.ac.uk/445733/},
year = {2021},
date = {2021-03-01},
journal = {Ad Hoc Networks},
volume = {112},
abstract = {The proliferation of ?Things? over a network creates the Internet of Things (IoT), where sensors integrate to collect data from the environment over long periods of time. The growth of IoT applications will inevitably involve co-locating multiple wireless sensor networks, each serving different applications with, possibly, different needs and constraints. Since energy is scarce in sensor nodes equipped with non-rechargeable batteries, energy harvesting technologies have been the focus of research in recent years. However, new problems arise as a result of their wide spatio-temporal variation. Such a shortcoming can be avoided if co-located networks cooperate with each other and share their available energy. Due to their unique characteristics and different owners, recently, we proposed a negotiation approach to deal with conflict of preferences. Unfortunately, negotiation can be impractical with a large number of participants, especially in an open environment. Given this, we introduce a new partner selection technique based on multi-armed bandits (MAB), that enables each node to learn the strategy that optimises its energy resources in the long term. Our results show that the proposed solution allows networks to repeatedly learn the current best energy partner in a dynamic environment. The performance of such a technique is evaluated through simulation and shows that a network can achieve an efficiency of 72% against the optimal strategy in the most challenging scenario studied in this work.},
keywords = {Agent-based sensor network, Automated negotiation, Energy management, Multi-armed bandit based learning, Reinforcement Learning, Wireless sensor networks},
pubstate = {published},
tppubtype = {article}
}
The proliferation of ?Things? over a network creates the Internet of Things (IoT), where sensors integrate to collect data from the environment over long periods of time. The growth of IoT applications will inevitably involve co-locating multiple wireless sensor networks, each serving different applications with, possibly, different needs and constraints. Since energy is scarce in sensor nodes equipped with non-rechargeable batteries, energy harvesting technologies have been the focus of research in recent years. However, new problems arise as a result of their wide spatio-temporal variation. Such a shortcoming can be avoided if co-located networks cooperate with each other and share their available energy. Due to their unique characteristics and different owners, recently, we proposed a negotiation approach to deal with conflict of preferences. Unfortunately, negotiation can be impractical with a large number of participants, especially in an open environment. Given this, we introduce a new partner selection technique based on multi-armed bandits (MAB), that enables each node to learn the strategy that optimises its energy resources in the long term. Our results show that the proposed solution allows networks to repeatedly learn the current best energy partner in a dynamic environment. The performance of such a technique is evaluated through simulation and shows that a network can achieve an efficiency of 72% against the optimal strategy in the most challenging scenario studied in this work.
Yazdanpanah, Vahid; Gerding, Enrico H; Stein, Sebastian; Dastani, Mehdi; Jonker, Catholijn M; Norman, Timothy J; Ramchurn, Sarvapali D
Responsibility ascription in trustworthy autonomous systems Proceedings Article
In: Embedding AI in Society (18/02/21 - 19/02/21), 2021.
Abstract | Links | BibTeX | Tags: Multiagent Systems, Reliable AI, Responsibility Reasoning, Trustworthy Autonomous Systems
@inproceedings{soton446459,
title = {Responsibility ascription in trustworthy autonomous systems},
author = {Vahid Yazdanpanah and Enrico H Gerding and Sebastian Stein and Mehdi Dastani and Catholijn M Jonker and Timothy J Norman and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/446459/},
year = {2021},
date = {2021-02-01},
booktitle = {Embedding AI in Society (18/02/21 - 19/02/21)},
abstract = {To develop and effectively deploy Trustworthy Autonomous Systems (TAS), we face various social, technological, legal, and ethical challenges in which different notions of responsibility can play a key role. In this work, we elaborate on these challenges, discuss research gaps, and show how the multidimensional notion of responsibility can play a key role to bridge them. We argue that TAS requires operational tools to represent and reason about the responsibilities of humans as well as AI agents.},
keywords = {Multiagent Systems, Reliable AI, Responsibility Reasoning, Trustworthy Autonomous Systems},
pubstate = {published},
tppubtype = {inproceedings}
}
To develop and effectively deploy Trustworthy Autonomous Systems (TAS), we face various social, technological, legal, and ethical challenges in which different notions of responsibility can play a key role. In this work, we elaborate on these challenges, discuss research gaps, and show how the multidimensional notion of responsibility can play a key role to bridge them. We argue that TAS requires operational tools to represent and reason about the responsibilities of humans as well as AI agents.
Beal, Ryan James; Middleton, Stuart; Norman, Timothy; Ramchurn, Sarvapali
Combining machine learning and human experts to predict match outcomes in football: A baseline model Proceedings Article
In: The Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (02/02/21 - 09/02/21), 2021.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Natural Language Processing
@inproceedings{soton445607,
title = {Combining machine learning and human experts to predict match outcomes in football: A baseline model},
author = {Ryan James Beal and Stuart Middleton and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/445607/},
year = {2021},
date = {2021-02-01},
booktitle = {The Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (02/02/21 - 09/02/21)},
abstract = {In this paper, we present a new application-focused benchmark dataset and results from a set of baseline Natural Language Processing and Machine Learning models for prediction of match outcomes for games of football (soccer). By doing so we give a baseline for the prediction accuracy that can be achieved exploiting both statistical match data and contextual articles from human sports journalists. Our dataset is focuses on a representative time-period over 6 seasons of the English Premier League, and includes newspaper match previews from The Guardian. The models presented in this paper achieve an accuracy of 63.18% showing a 6.9% boost on the traditional statistical methods.},
keywords = {Artificial Intelligence, Natural Language Processing},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we present a new application-focused benchmark dataset and results from a set of baseline Natural Language Processing and Machine Learning models for prediction of match outcomes for games of football (soccer). By doing so we give a baseline for the prediction accuracy that can be achieved exploiting both statistical match data and contextual articles from human sports journalists. Our dataset is focuses on a representative time-period over 6 seasons of the English Premier League, and includes newspaper match previews from The Guardian. The models presented in this paper achieve an accuracy of 63.18% showing a 6.9% boost on the traditional statistical methods.
Lhopital, Sacha; Aknine, Samir; Ramchurn, Sarvapali; Thavonekham, Vincent; Vu, Huan
Decentralised control of intelligent devices: a healthcare facility study Proceedings Article
In: Bassiliades, Nick; Chalkiadakis, Georgios; Jonge, Dave (Ed.): Multi-Agent Systems and Agreement Technologies - 17th European Conference, EUMAS 2020, and 7th International Conference, AT 2020, Revised Selected Papers, pp. 20–36, Springer, 2021.
Abstract | Links | BibTeX | Tags: DCOP, DPOP, Healthcare, IoT
@inproceedings{soton447983,
title = {Decentralised control of intelligent devices: a healthcare facility study},
author = {Sacha Lhopital and Samir Aknine and Sarvapali Ramchurn and Vincent Thavonekham and Huan Vu},
editor = {Nick Bassiliades and Georgios Chalkiadakis and Dave Jonge},
url = {https://eprints.soton.ac.uk/447983/},
year = {2021},
date = {2021-01-01},
booktitle = {Multi-Agent Systems and Agreement Technologies - 17th European Conference, EUMAS 2020, and 7th International Conference, AT 2020, Revised Selected Papers},
volume = {12520 LNAI},
pages = {20–36},
publisher = {Springer},
abstract = {ensuremath<pensuremath>We present a novel approach to the management of notifications from devices in a healthcare setting. We employ a distributed constraint optimisation (DCOP) approach to the delivery of notification for healthcare assistants that aims to preserve the privacy of patients while reducing the intrusiveness of such notifications. Our approach reduces the workload of the assistants and improves patient safety by automating task allocation while ensuring high priority needs are addressed in a timely manner. We propose and evaluate several DCOP models both in simulation and in real-world deployments. Our models are shown to be efficient both in terms of computation and communication costs.ensuremath</pensuremath>},
keywords = {DCOP, DPOP, Healthcare, IoT},
pubstate = {published},
tppubtype = {inproceedings}
}
ensuremath<pensuremath>We present a novel approach to the management of notifications from devices in a healthcare setting. We employ a distributed constraint optimisation (DCOP) approach to the delivery of notification for healthcare assistants that aims to preserve the privacy of patients while reducing the intrusiveness of such notifications. Our approach reduces the workload of the assistants and improves patient safety by automating task allocation while ensuring high priority needs are addressed in a timely manner. We propose and evaluate several DCOP models both in simulation and in real-world deployments. Our models are shown to be efficient both in terms of computation and communication costs.ensuremath</pensuremath>
Ryan, James Beal; Chalkiadakis, Georgios; Norman, Timothy; Ramchurn, Sarvapali
Optimising long-term outcomes using real-world fluent objectives: an application to football Proceedings Article
In: 20th International Conference on Autonomous Agents and Multiagent Systems (03/05/21 - 07/05/21), pp. 196–204, 2021.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton449655,
title = {Optimising long-term outcomes using real-world fluent objectives: an application to football},
author = {James Beal Ryan and Georgios Chalkiadakis and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449655/},
year = {2021},
date = {2021-01-01},
booktitle = {20th International Conference on Autonomous Agents and Multiagent Systems (03/05/21 - 07/05/21)},
pages = {196–204},
abstract = {In this paper, we present a novel approach for optimising long-term tactical and strategic decision-making in football (soccer) by encapsulating events in a league environment across a given time frame. ensuremath<br/ensuremath>We model the teams' objectives for a season and track how these evolve as games unfold to give a fluent objective that can aid in decision-making games. We develop Markov chain Monte Carlo and deep learning-based algorithms that make use of the fluent objectives in order to learn from prior games and other games in the environment and increase the teams? long-term performance. Simulations of our approach using real-world datasets from 760 matches shows that by using optimised tactics with our fluent objective and prior games, we can on average increase teams mean expected finishing distribution in the league by up to 35.6%.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we present a novel approach for optimising long-term tactical and strategic decision-making in football (soccer) by encapsulating events in a league environment across a given time frame. ensuremath<br/ensuremath>We model the teams' objectives for a season and track how these evolve as games unfold to give a fluent objective that can aid in decision-making games. We develop Markov chain Monte Carlo and deep learning-based algorithms that make use of the fluent objectives in order to learn from prior games and other games in the environment and increase the teams? long-term performance. Simulations of our approach using real-world datasets from 760 matches shows that by using optimised tactics with our fluent objective and prior games, we can on average increase teams mean expected finishing distribution in the league by up to 35.6%.
Merhej, Charbel; Ryan, James Beal; Matthews, Tim; Ramchurn, Sarvapali
What happened next? Using deep learning to value defensive actions in football event-data Proceedings Article
In: KDD 2021 (14/08/21 - 18/08/21), pp. 3394–3403, 2021.
Abstract | Links | BibTeX | Tags: applied machine learning, deep learning, defensive actions, football, neural networks, sports analytics
@inproceedings{soton449656,
title = {What happened next? Using deep learning to value defensive actions in football event-data},
author = {Charbel Merhej and James Beal Ryan and Tim Matthews and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449656/},
year = {2021},
date = {2021-01-01},
booktitle = {KDD 2021 (14/08/21 - 18/08/21)},
pages = {3394–3403},
abstract = {Objectively quantifying the value of player actions in football (soccer) is a challenging problem. To date, studies in football analytics have mainly focused on the attacking side of the game, while there has been less work on event-driven metrics for valuing defensive actions (e.g., tackles and interceptions). Therefore in this paper, we use deep learning techniques to define a novel metric that values such defensive actions by studying the threat of passages of play that preceded them. By doing so, we are able to value defensive actions based on what they prevented from happening in the game. Our Defensive Action Expected Threat (DAxT) model has been validated using real-world event-data from the 2017/2018 and 2018/2019 English Premier League seasons, and we combine our model outputs with additional features to derive an overall rating of defensive ability for players. Overall, we find that our model is able to predict the impact of defensive actions allowing us to better value defenders using event-data.},
keywords = {applied machine learning, deep learning, defensive actions, football, neural networks, sports analytics},
pubstate = {published},
tppubtype = {inproceedings}
}
Objectively quantifying the value of player actions in football (soccer) is a challenging problem. To date, studies in football analytics have mainly focused on the attacking side of the game, while there has been less work on event-driven metrics for valuing defensive actions (e.g., tackles and interceptions). Therefore in this paper, we use deep learning techniques to define a novel metric that values such defensive actions by studying the threat of passages of play that preceded them. By doing so, we are able to value defensive actions based on what they prevented from happening in the game. Our Defensive Action Expected Threat (DAxT) model has been validated using real-world event-data from the 2017/2018 and 2018/2019 English Premier League seasons, and we combine our model outputs with additional features to derive an overall rating of defensive ability for players. Overall, we find that our model is able to predict the impact of defensive actions allowing us to better value defenders using event-data.
2020
Beal, Ryan James; Norman, Timothy; Ramchurn, Sarvapali
Optimising daily fantasy sports teams with artificial intelligence Journal Article
In: International Journal of Computer Science in Sport, vol. 19, no. 2, 2020.
Abstract | Links | BibTeX | Tags:
@article{soton445995,
title = {Optimising daily fantasy sports teams with artificial intelligence},
author = {Ryan James Beal and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/445995/},
year = {2020},
date = {2020-12-01},
journal = {International Journal of Computer Science in Sport},
volume = {19},
number = {2},
abstract = {This paper outlines a novel approach to optimising teams for Daily Fantasy Sports (DFS) contests. To this end, we propose a number of new models and algorithms to solve the team formation problems posed by DFS. Specifically, we focus on the National Football League (NFL) and predict the performance of real-world players to form the optimal fantasy team using mixed-integer programming. We test our solutions using real-world data-sets from across four seasons (2014-2017). We highlight the advantage that can be gained from using our machine-based methods and show that our solutions outperform existing benchmarks, turning a profit in up to 81.3% of DFS game-weeks over a season},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper outlines a novel approach to optimising teams for Daily Fantasy Sports (DFS) contests. To this end, we propose a number of new models and algorithms to solve the team formation problems posed by DFS. Specifically, we focus on the National Football League (NFL) and predict the performance of real-world players to form the optimal fantasy team using mixed-integer programming. We test our solutions using real-world data-sets from across four seasons (2014-2017). We highlight the advantage that can be gained from using our machine-based methods and show that our solutions outperform existing benchmarks, turning a profit in up to 81.3% of DFS game-weeks over a season
Beal, Ryan James; Norman, Timothy; Ramchurn, Sarvapali
A critical comparison of machine learning classifiers to predict match outcomes in the NFL Journal Article
In: International Journal of Computer Science in Sport, vol. 19, no. 2, 2020.
Abstract | Links | BibTeX | Tags:
@article{soton446078,
title = {A critical comparison of machine learning classifiers to predict match outcomes in the NFL},
author = {Ryan James Beal and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/446078/},
year = {2020},
date = {2020-12-01},
journal = {International Journal of Computer Science in Sport},
volume = {19},
number = {2},
abstract = {In this paper, we critically evaluate the performance of nine machine learning classification techniques when applied to the match outcome prediction problem presented by American Football. Specifically, we implement and test nine techniques using real-world datasets of 1280 games over 5 seasons from the National Football League (NFL). We test the nine different classifier techniques using a total of 42 features for each team and we find that the best performing algorithms are able to improve one previous published works. The algoriothms achieve an accuracy of between 44.64% for a Guassian Process classifier to 67.53% with a Na"ive Bayes classifer. We also test each classifier on a year by year basis and compare our results to those of the bookmakers and other leading academic papers.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper, we critically evaluate the performance of nine machine learning classification techniques when applied to the match outcome prediction problem presented by American Football. Specifically, we implement and test nine techniques using real-world datasets of 1280 games over 5 seasons from the National Football League (NFL). We test the nine different classifier techniques using a total of 42 features for each team and we find that the best performing algorithms are able to improve one previous published works. The algoriothms achieve an accuracy of between 44.64% for a Guassian Process classifier to 67.53% with a Na"ive Bayes classifer. We also test each classifier on a year by year basis and compare our results to those of the bookmakers and other leading academic papers.
Rigas, Emmanouil S; Gerding, Enrico; Stein, Sebastian; Ramchurn, Sarvapali D; Bassiliades, Nick
Mechanism design for efficient allocation of electric vehicles to charging stations Proceedings Article
In: SETN 2020: 11th Hellenic Conference on Artificial Intelligence, pp. 10–15, 2020.
Abstract | Links | BibTeX | Tags:
@inproceedings{soton446412,
title = {Mechanism design for efficient allocation of electric vehicles to charging stations},
author = {Emmanouil S Rigas and Enrico Gerding and Sebastian Stein and Sarvapali D Ramchurn and Nick Bassiliades},
url = {https://eprints.soton.ac.uk/446412/},
year = {2020},
date = {2020-09-01},
booktitle = {SETN 2020: 11th Hellenic Conference on Artificial Intelligence},
pages = {10–15},
abstract = {The electrification of transport can significantly reduce CO2 emissions and their negative impact on the environment. In this paper, we study the problem of allocating Electric Vehicles (EVs) to charging stations and scheduling their charging. We develop an offline solution that treats EV users as self-interested agents that aim to maximise their profit and minimise the impact on their schedule. We formulate the problem of the optimal EV to charging station allocation as a Mixed Integer Programming (MIP) one and we propose two pricing mechanisms: A fixed-price one, and another that is based on the well known Vickrey-Clark-Groves (VCG) mechanism. We observe that the VCG mechanism services on average 1.5% more EVs than the fixed-price one. In addition, when the stations get congested, VCG leads to higher prices for the EVs and higher profit for the stations, but lower utility for the EVs. However, the VCG mechanism guarantees truthful reporting of the EVs? preferences.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The electrification of transport can significantly reduce CO2 emissions and their negative impact on the environment. In this paper, we study the problem of allocating Electric Vehicles (EVs) to charging stations and scheduling their charging. We develop an offline solution that treats EV users as self-interested agents that aim to maximise their profit and minimise the impact on their schedule. We formulate the problem of the optimal EV to charging station allocation as a Mixed Integer Programming (MIP) one and we propose two pricing mechanisms: A fixed-price one, and another that is based on the well known Vickrey-Clark-Groves (VCG) mechanism. We observe that the VCG mechanism services on average 1.5% more EVs than the fixed-price one. In addition, when the stations get congested, VCG leads to higher prices for the EVs and higher profit for the stations, but lower utility for the EVs. However, the VCG mechanism guarantees truthful reporting of the EVs? preferences.
Oluwasuji, Olabambo Ifeoluwa; Malik, Obaid; Zhang, Jie; Ramchurn, Sarvapali Dyanand
Solving the fair electric load shedding problem in developing countries Journal Article
In: Autonomous Agents and Multi-Agent Systems, vol. 34, no. 1, pp. 12, 2020.
BibTeX | Tags:
@article{oluwasuji2020solving,
title = {Solving the fair electric load shedding problem in developing countries},
author = {Olabambo Ifeoluwa Oluwasuji and Obaid Malik and Jie Zhang and Sarvapali Dyanand Ramchurn},
year = {2020},
date = {2020-01-01},
journal = {Autonomous Agents and Multi-Agent Systems},
volume = {34},
number = {1},
pages = {12},
publisher = {Springer},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Oluwasuji, Olabambo Ifeoluwa; Malik, Obaid; Zhang, Jie; Ramchurn, Sarvapali Dyanand
Solving the fair electric load shedding problem in developing countries Journal Article
In: Auton. Agents Multi Agent Syst., vol. 34, no. 1, pp. 12, 2020.
@article{DBLP:journals/aamas/OluwasujiMZR20,
title = {Solving the fair electric load shedding problem in developing countries},
author = {Olabambo Ifeoluwa Oluwasuji and Obaid Malik and Jie Zhang and Sarvapali Dyanand Ramchurn},
url = {https://doi.org/10.1007/s10458-019-09428-8},
doi = {10.1007/s10458-019-09428-8},
year = {2020},
date = {2020-01-01},
journal = {Auton. Agents Multi Agent Syst.},
volume = {34},
number = {1},
pages = {12},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
1.
Deshmukh, Jayati; Liang, Zijie; Yazdanpanah, Vahid; Stein, Sebastian; Ramchurn, Sarvalpali D.
Serious games for ethical preference elicitation Proceedings Article
In: AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25), 2025.
@inproceedings{soton498743,
title = {Serious games for ethical preference elicitation},
author = {Jayati Deshmukh and Zijie Liang and Vahid Yazdanpanah and Sebastian Stein and Sarvalpali D. Ramchurn},
url = {https://eprints.soton.ac.uk/498743/},
year = {2025},
date = {2025-05-01},
booktitle = {AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25)},
abstract = {Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.
2.
Thavanesan, Navamayooran; Farahi, Arya; Parfitt, Charlotte; Belkhatir, Zehor; Azim, Tayyaba; Vallejos, Elvira Perez; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Insights from explainable AI in oesophageal cancer team decisions Journal Article
In: Computers in Biology and Medicine, vol. 180, 2024, (For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.).
@article{soton493238,
title = {Insights from explainable AI in oesophageal cancer team decisions},
author = {Navamayooran Thavanesan and Arya Farahi and Charlotte Parfitt and Zehor Belkhatir and Tayyaba Azim and Elvira Perez Vallejos and Zoë Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/493238/},
year = {2024},
date = {2024-08-01},
journal = {Computers in Biology and Medicine},
volume = {180},
abstract = {ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>},
note = {For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>
3.
Naiseh, Mohammad; Webb, Catherine; Underwood, Tim; Ramchurn, Gopal; Walters, Zoe; Thavanesan, Navamayooran; Vigneswaran, Ganesh
XAI for group-AI interaction: towards collaborative and inclusive explanations Proceedings Article
In: Longo, Luca; Liu, Weiru; Montavon, Gregoire (Ed.): Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024), pp. 249–256, CEUR Workshop Proceedings, 2024.
@inproceedings{soton497829,
title = {XAI for group-AI interaction: towards collaborative and inclusive explanations},
author = {Mohammad Naiseh and Catherine Webb and Tim Underwood and Gopal Ramchurn and Zoe Walters and Navamayooran Thavanesan and Ganesh Vigneswaran},
editor = {Luca Longo and Weiru Liu and Gregoire Montavon},
url = {https://eprints.soton.ac.uk/497829/},
year = {2024},
date = {2024-07-01},
booktitle = {Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024)},
volume = {3793},
pages = {249–256},
publisher = {CEUR Workshop Proceedings},
abstract = {ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>
4.
Early, Joseph Arthur
Interpretable multiple instance learning PhD Thesis
University of Southampton, 2024.
@phdthesis{soton490767,
title = {Interpretable multiple instance learning},
author = {Joseph Arthur Early},
url = {https://eprints.soton.ac.uk/490767/},
year = {2024},
date = {2024-06-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>
5.
Kiden, Sarah; Stahl, Bernd; Townsend, Beverley; Maple, Carsten; Vincent, Charles; Sampson, Fraser; Gilbert, Geoff; Smith, Helen; Deshmukh, Jayati; Ross, Jen; Williams, Jennifer; Rincon, Jesus Martinez; Lisinska, Justyna; O?Shea, Karen; Abreu, Márjory Da Costa; Bencomo, Nelly; Deb, Oishi; Winter, Peter; Li, Phoebe; Torr, Philip; Lau, Pin Lean; Iniesta, Raquel; Ramchurn, Gopal; Stein, Sebastian; Yazdanpanah, Vahid
Responsible AI governance: A response to UN interim report on governing AI for humanity Technical Report
no. 10.5258/SOTON/PP0057, 2024.
@techreport{soton488908,
title = {Responsible AI governance: A response to UN interim report on governing AI for humanity},
author = {Sarah Kiden and Bernd Stahl and Beverley Townsend and Carsten Maple and Charles Vincent and Fraser Sampson and Geoff Gilbert and Helen Smith and Jayati Deshmukh and Jen Ross and Jennifer Williams and Jesus Martinez Rincon and Justyna Lisinska and Karen O?Shea and Márjory Da Costa Abreu and Nelly Bencomo and Oishi Deb and Peter Winter and Phoebe Li and Philip Torr and Pin Lean Lau and Raquel Iniesta and Gopal Ramchurn and Sebastian Stein and Vahid Yazdanpanah},
url = {https://eprints.soton.ac.uk/488908/},
year = {2024},
date = {2024-03-01},
number = {10.5258/SOTON/PP0057},
publisher = {Public Policy, University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
6.
Abioye, Ayodeji O.; Hunt, William; Gu, Yue; Schneiders, Eike; Naiseh, Mohammad; Fischer, Joel E.; Ramchurn, Sarvapali D.; Soorati, Mohammad D.; Archibald, Blair; Sevegnani, Michele
The effect of predictive formal modelling at runtime on performance in human-swarm interaction Proceedings Article
In: HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction, pp. 172?176, Association for Computing Machinery, 2024, (Publisher Copyright: © 2024 Copyright held by the owner/author(s)).
@inproceedings{soton488273,
title = {The effect of predictive formal modelling at runtime on performance in human-swarm interaction},
author = {Ayodeji O. Abioye and William Hunt and Yue Gu and Eike Schneiders and Mohammad Naiseh and Joel E. Fischer and Sarvapali D. Ramchurn and Mohammad D. Soorati and Blair Archibald and Michele Sevegnani},
url = {https://eprints.soton.ac.uk/488273/},
year = {2024},
date = {2024-03-01},
booktitle = {HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction},
pages = {172?176},
publisher = {Association for Computing Machinery},
abstract = {Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.},
note = {Publisher Copyright:
© 2024 Copyright held by the owner/author(s)},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.
7.
Soorati, Mohammad D.; Naiseh, Mohammad; Hunt, William; Parnell, Katie; Clark, Jediah; Ramchurn, Sarvapali D.
Enabling trustworthiness in human-swarm systems through a digital twin Book Section
In: Dasgupta, Prithviraj; Llinas, James; Gillespie, Tony; Fouse, Scott; Lawless, William; Mittu, Ranjeev; Sofge, Donlad (Ed.): Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams, pp. 93–125, Academic Press, 2024, (Publisher Copyright: © 2024 Elsevier Inc. All rights reserved.).
@incollection{soton491769,
title = {Enabling trustworthiness in human-swarm systems through a digital twin},
author = {Mohammad D. Soorati and Mohammad Naiseh and William Hunt and Katie Parnell and Jediah Clark and Sarvapali D. Ramchurn},
editor = {Prithviraj Dasgupta and James Llinas and Tony Gillespie and Scott Fouse and William Lawless and Ranjeev Mittu and Donlad Sofge},
url = {https://eprints.soton.ac.uk/491769/},
year = {2024},
date = {2024-02-01},
booktitle = {Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams},
pages = {93–125},
publisher = {Academic Press},
abstract = {Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.},
note = {Publisher Copyright:
© 2024 Elsevier Inc. All rights reserved.},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.
8.
Thavanesan, Navamayooran; Parfitt, Charlotte; Bodala, Indu; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy; Vigneswaran, Ganesh
Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction Miscellaneous
2024.
@misc{soton497828,
title = {Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction},
author = {Navamayooran Thavanesan and Charlotte Parfitt and Indu Bodala and Zoë Walters and Sarvapali Ramchurn and Timothy Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/497828/},
year = {2024},
date = {2024-01-01},
journal = {European Journal of Surgical Oncology},
volume = {50},
number = {1},
abstract = {Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035
9.
Kelly, Thomas Graham; Soorati, Mohammad; Zauner, Klaus-Peter; Ramchurn, Gopal; Tarapore, Danesh
Trade-offs of dynamic control structure in human-swarm systems Proceedings Article
In: The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024, 2024.
@inproceedings{soton492838,
title = {Trade-offs of dynamic control structure in human-swarm systems},
author = {Thomas Graham Kelly and Mohammad Soorati and Klaus-Peter Zauner and Gopal Ramchurn and Danesh Tarapore},
url = {https://eprints.soton.ac.uk/492838/},
year = {2024},
date = {2024-01-01},
booktitle = {The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024},
abstract = {Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.
10.
Early, Joseph; Deweese, Ying-Jung Chen; Evers, Christine; Ramchurn, Sarvapali
Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation Journal Article
In: Environmental Data Science, vol. 2, pp. 18, 2023.
@article{soton490766,
title = {Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation},
author = {Joseph Early and Ying-Jung Chen Deweese and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/490766/},
year = {2023},
date = {2023-12-01},
journal = {Environmental Data Science},
volume = {2},
pages = {18},
abstract = {Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.
11.
Rigley, Eryn; Bentley, Caitlin; Krook, Joshua; Ramchurn, Gopal
Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries Journal Article
In: Global Policy, 2023, (Funding Information: This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.).
@article{soton485727,
title = {Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries},
author = {Eryn Rigley and Caitlin Bentley and Joshua Krook and Gopal Ramchurn},
url = {https://eprints.soton.ac.uk/485727/},
year = {2023},
date = {2023-12-01},
journal = {Global Policy},
abstract = {ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>},
note = {Funding Information:
This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>
12.
Singh, Lokesh; Ramchurn, Gopal
The effect of automated agents on individual performance under induced stress Proceedings Article
In: Kalra, Jay (Ed.): Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition), pp. 118–127, AHFE International, 2023.
@inproceedings{soton485655,
title = {The effect of automated agents on individual performance under induced stress},
author = {Lokesh Singh and Gopal Ramchurn},
editor = {Jay Kalra},
url = {https://eprints.soton.ac.uk/485655/},
year = {2023},
date = {2023-11-01},
booktitle = {Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition)},
pages = {118–127},
publisher = {AHFE International},
abstract = {Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.
13.
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, vol. 49, no. 11, 2023, (Publisher Copyright: © 2023 The Author(s)).
@article{soton479497b,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-11-01},
journal = {European Journal of Surgical Oncology},
volume = {49},
number = {11},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
© 2023 The Author(s)},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
14.
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub Response Miscellaneous
2023.
@misc{soton481740,
title = {AI large language models inquiry: TASHub Response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
15.
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub response Miscellaneous
2023.
@misc{soton481740b,
title = {AI large language models inquiry: TASHub response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
16.
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, 2023, (Publisher Copyright: copyright 2023 The Author(s)).
@article{soton479497,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-07-01},
journal = {European Journal of Surgical Oncology},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
copyright 2023 The Author(s)},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
17.
Abioye, Ayodeji
University of Southampton, 2023.
@phdthesis{soton479472,
title = {Multimodal speech and visual gesture control interface technique for small unmanned multirotor aircraft},
author = {Ayodeji Abioye},
url = {https://eprints.soton.ac.uk/479472/},
year = {2023},
date = {2023-07-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>
18.
Abioye, Ayodeji; Naiseh, Mohammad; Hunt, William; Clark, Jediah R; Ramchurn, Sarvapali D; Soorati, Mohammad
The effect of data visualisation quality and task density on human-swarm interaction Proceedings Article
In: Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), IEEE, 2023.
@inproceedings{soton479970,
title = {The effect of data visualisation quality and task density on human-swarm interaction},
author = {Ayodeji Abioye and Mohammad Naiseh and William Hunt and Jediah R Clark and Sarvapali D Ramchurn and Mohammad Soorati},
url = {https://eprints.soton.ac.uk/479970/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
booktitle = {Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
publisher = {IEEE},
abstract = {Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.
19.
Krook, Joshua; McAuley, Derek; Anderson, Stuart; Downer, John; Winter, Peter; Ramchurn, Sarvapali D
AI Foundation Models: initial review, CMA Consultation, TAS Hub Response Miscellaneous
2023.
@misc{soton477553,
title = {AI Foundation Models: initial review, CMA Consultation, TAS Hub Response},
author = {Joshua Krook and Derek McAuley and Stuart Anderson and John Downer and Peter Winter and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477553/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
publisher = {University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
20.
Krook, Joshua; Downer, John; Winter, Peter; Williams, Jennifer; Ives, Jonathan; Bratu, Roxana; Sheir, Stephanie; Williams, Robin; Anderson, Stuart; Li, Phoebe; Ramamoorthy, Subramanian; Ramchurn, Sarvapali
AI regulation: a pro-innovation approach ? policy proposals: TASHub Response Miscellaneous
2023.
@misc{soton478329,
title = {AI regulation: a pro-innovation approach ? policy proposals: TASHub Response},
author = {Joshua Krook and John Downer and Peter Winter and Jennifer Williams and Jonathan Ives and Roxana Bratu and Stephanie Sheir and Robin Williams and Stuart Anderson and Phoebe Li and Subramanian Ramamoorthy and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478329/},
year = {2023},
date = {2023-06-01},
publisher = {University of Southampton},
abstract = {Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence
Selected Publications
Multi-agent signal-less intersection management with dynamic platoon formation
No data available.
https://eprints.soton.ac.uk/478647/
@inproceedings{soton478647,
title = {Multi-agent signal-less intersection management with dynamic platoon formation},
author = {Phuriwat Worrawichaipat and Enrico Gerding and Ioannis Kaparias and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478647/},
year = {2023},
date = {2023-05-01},
booktitle = {22nd International Conference on Autonomous Agents and Multiagent Systems (29/05/23 - 02/06/23)},
pages = {1542--1550},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Multi-agent signal-less intersection management with dynamic platoon formation},
author = {Phuriwat Worrawichaipat and Enrico Gerding and Ioannis Kaparias and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478647/},
year = {2023},
date = {2023-05-01},
booktitle = {22nd International Conference on Autonomous Agents and Multiagent Systems (29/05/23 - 02/06/23)},
pages = {1542--1550},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
AI Foundation Models: initial review, CMA Consultation, TAS Hub Response
No data available.
https://eprints.soton.ac.uk/477553/
@misc{soton477553,
title = {AI Foundation Models: initial review, CMA Consultation, TAS Hub Response},
author = {Joshua Krook and Derek McAuley and Stuart Anderson and John Downer and Peter Winter and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477553/},
year = {2023},
date = {2023-06-01},
publisher = {University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
title = {AI Foundation Models: initial review, CMA Consultation, TAS Hub Response},
author = {Joshua Krook and Derek McAuley and Stuart Anderson and John Downer and Peter Winter and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477553/},
year = {2023},
date = {2023-06-01},
publisher = {University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
The effect of data visualisation quality and task density on human-swarm interaction
No data available.
https://eprints.soton.ac.uk/479970/
@inproceedings{soton479970,
title = {The effect of data visualisation quality and task density on human-swarm interaction},
author = {Ayodeji Abioye and Mohammad Naiseh and William Hunt and Jediah R Clark and Sarvapali D Ramchurn and Mohammad Soorati},
url = {https://eprints.soton.ac.uk/479970/},
year = {2023},
date = {2023-06-01},
booktitle = {Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
publisher = {IEEE},
abstract = {Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {The effect of data visualisation quality and task density on human-swarm interaction},
author = {Ayodeji Abioye and Mohammad Naiseh and William Hunt and Jediah R Clark and Sarvapali D Ramchurn and Mohammad Soorati},
url = {https://eprints.soton.ac.uk/479970/},
year = {2023},
date = {2023-06-01},
booktitle = {Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
publisher = {IEEE},
abstract = {Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Demonstrating performance benefits of human-swarm teaming
Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.
https://eprints.soton.ac.uk/479903/
@inproceedings{soton479903,
title = {Demonstrating performance benefits of human-swarm teaming},
author = {William Hunt and Jack Ryan and Ayodeji O Abioye and Sarvapali D Ramchurn and Mohammad D Soorati},
url = {https://eprints.soton.ac.uk/479903/},
year = {2023},
date = {2023-05-01},
booktitle = {Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems},
pages = {3062--3064},
publisher = {International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS)},
abstract = {Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Demonstrating performance benefits of human-swarm teaming},
author = {William Hunt and Jack Ryan and Ayodeji O Abioye and Sarvapali D Ramchurn and Mohammad D Soorati},
url = {https://eprints.soton.ac.uk/479903/},
year = {2023},
date = {2023-05-01},
booktitle = {Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems},
pages = {3062--3064},
publisher = {International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS)},
abstract = {Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
2025
Deshmukh, Jayati; Liang, Zijie; Yazdanpanah, Vahid; Stein, Sebastian; Ramchurn, Sarvalpali D.
Serious games for ethical preference elicitation Proceedings Article
In: AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25), 2025.
@inproceedings{soton498743,
title = {Serious games for ethical preference elicitation},
author = {Jayati Deshmukh and Zijie Liang and Vahid Yazdanpanah and Sebastian Stein and Sarvalpali D. Ramchurn},
url = {https://eprints.soton.ac.uk/498743/},
year = {2025},
date = {2025-05-01},
booktitle = {AAMAS - 2025 : The 24th International Conference on Autonomous Agents and Multiagent Systems (19/05/25 - 23/05/25)},
abstract = {Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Autonomous agents acting on behalf of humans must act according to their ethical preferences. However, ethical preferences are latent and abstract and thus it is challenging to elicit them. To address this, we present a serious game that helps elicit ethical preferences in a more dynamic and engaging way than traditional methods such as questionnaires or simple dilemmas.
2024
Thavanesan, Navamayooran; Farahi, Arya; Parfitt, Charlotte; Belkhatir, Zehor; Azim, Tayyaba; Vallejos, Elvira Perez; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Insights from explainable AI in oesophageal cancer team decisions Journal Article
In: Computers in Biology and Medicine, vol. 180, 2024, (For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.).
@article{soton493238,
title = {Insights from explainable AI in oesophageal cancer team decisions},
author = {Navamayooran Thavanesan and Arya Farahi and Charlotte Parfitt and Zehor Belkhatir and Tayyaba Azim and Elvira Perez Vallejos and Zoë Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/493238/},
year = {2024},
date = {2024-08-01},
journal = {Computers in Biology and Medicine},
volume = {180},
abstract = {ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>},
note = {For the purpose of open access, the authors have applied a Creative Commons attribution license (CC-BY) to any Author Accepted Manuscript version arising from this submission.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT).ensuremath</pensuremath>ensuremath<pensuremath>Methods: retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic.�ensuremath</pensuremath>ensuremath<pensuremath>Results: amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75?85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age.�ensuremath</pensuremath>ensuremath<pensuremath>Conclusion: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.ensuremath</pensuremath>
Naiseh, Mohammad; Webb, Catherine; Underwood, Tim; Ramchurn, Gopal; Walters, Zoe; Thavanesan, Navamayooran; Vigneswaran, Ganesh
XAI for group-AI interaction: towards collaborative and inclusive explanations Proceedings Article
In: Longo, Luca; Liu, Weiru; Montavon, Gregoire (Ed.): Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024), pp. 249–256, CEUR Workshop Proceedings, 2024.
@inproceedings{soton497829,
title = {XAI for group-AI interaction: towards collaborative and inclusive explanations},
author = {Mohammad Naiseh and Catherine Webb and Tim Underwood and Gopal Ramchurn and Zoe Walters and Navamayooran Thavanesan and Ganesh Vigneswaran},
editor = {Luca Longo and Weiru Liu and Gregoire Montavon},
url = {https://eprints.soton.ac.uk/497829/},
year = {2024},
date = {2024-07-01},
booktitle = {Joint Proceedings of the xAI 2024 Late-breaking Work, Demos and Doctoral Consortium co-located with the 2nd World Conference on eXplainable Artificial Intelligence (xAI 2024)},
volume = {3793},
pages = {249–256},
publisher = {CEUR Workshop Proceedings},
abstract = {ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
ensuremath<pensuremath>The increasing integration of Machine Learning (ML) into decision-making across various sectors has raised concerns about ethics, legality, explainability, and safety, highlighting the necessity of human oversight. In response, eXplainable AI (XAI) has emerged as a means to enhance transparency by providing insights into ML model decisions and offering humans an understanding of the underlying logic. Despite its potential, existing XAI models often lack practical usability and fail to improve human-AI performance, as they may introduce issues such as overreliance. This underscores the need for further research in Human-Centered XAI to improve the usability of current XAI methods. Notably, much of the current research focuses on one-to-one interactions between the XAI and individual decision-makers, overlooking the dynamics of many-to-one relationships in real-world scenarios where groups of humans collaborate using XAI in collective decision-making. In this late-breaking work, we draw upon current work in Human-Centered XAI research and discuss how XAI design could be transitioned to group-AI interaction. We discuss four potential challenges in the transition of XAI from human-AI interaction to group-AI interaction. This paper contributes to advancing the field of Human-Centered XAI and facilitates the discussion on group-XAI interaction, calling for further research in this area.ensuremath</pensuremath>
Early, Joseph Arthur
Interpretable multiple instance learning PhD Thesis
University of Southampton, 2024.
@phdthesis{soton490767,
title = {Interpretable multiple instance learning},
author = {Joseph Arthur Early},
url = {https://eprints.soton.ac.uk/490767/},
year = {2024},
date = {2024-06-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
With the rising use of Artificial Intelligence (AI) and Machine Learning (ML) methods, there comes an increasing need to understand how automated systems make decisions. Interpretable ML provides insight into the underlying reasoning behind AI and ML models while not stifling their predictive performance. Doing so is important for many reasons, such as facilitating trust, increasing transparency, and providing improved collaboration and control through a better understanding of automated decision-making. Interpretability is very relevant across many ML paradigms and application domains. Multiple Instance Learning (MIL) is an ML paradigm where data are grouped into bags of instances, and only the bags are labelled (rather than each instance). This is beneficial in alleviating expensive labelling procedures and can be used to exploit the underlying structure of data. This thesis investigates how interpretability can be achieved within MIL. It begins with a formalisation of interpretable MIL, and then proposes a suite of model-agnostic post-hoc methods. This work is then extended to the specific application domain of high-resolution satellite imagery, using novel inherently interpretable MIL approaches that operate at multiple resolutions. Following on from work in the vision domain, new methods for interpretable MIL are developed for sequential data. First, it is explored in the domain of Reward Modelling (RM) for Reinforcement Learning (RL), demonstrating that interpretable MIL can be used to not only understand a model but also improve its predictive performance. This is mirrored in the application of interpretable MIL to Time Series Classification (TSC), where it is integrated into state-of-the-art methods and is able to improve both their interpretability and predictive performance. The integration into existing models to provide inherent interpretability means these benefits are delivered with little additional computational cost. ensuremath<br/ensuremath>
Kiden, Sarah; Stahl, Bernd; Townsend, Beverley; Maple, Carsten; Vincent, Charles; Sampson, Fraser; Gilbert, Geoff; Smith, Helen; Deshmukh, Jayati; Ross, Jen; Williams, Jennifer; Rincon, Jesus Martinez; Lisinska, Justyna; O?Shea, Karen; Abreu, Márjory Da Costa; Bencomo, Nelly; Deb, Oishi; Winter, Peter; Li, Phoebe; Torr, Philip; Lau, Pin Lean; Iniesta, Raquel; Ramchurn, Gopal; Stein, Sebastian; Yazdanpanah, Vahid
Responsible AI governance: A response to UN interim report on governing AI for humanity Technical Report
no. 10.5258/SOTON/PP0057, 2024.
@techreport{soton488908,
title = {Responsible AI governance: A response to UN interim report on governing AI for humanity},
author = {Sarah Kiden and Bernd Stahl and Beverley Townsend and Carsten Maple and Charles Vincent and Fraser Sampson and Geoff Gilbert and Helen Smith and Jayati Deshmukh and Jen Ross and Jennifer Williams and Jesus Martinez Rincon and Justyna Lisinska and Karen O?Shea and Márjory Da Costa Abreu and Nelly Bencomo and Oishi Deb and Peter Winter and Phoebe Li and Philip Torr and Pin Lean Lau and Raquel Iniesta and Gopal Ramchurn and Sebastian Stein and Vahid Yazdanpanah},
url = {https://eprints.soton.ac.uk/488908/},
year = {2024},
date = {2024-03-01},
number = {10.5258/SOTON/PP0057},
publisher = {Public Policy, University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Abioye, Ayodeji O.; Hunt, William; Gu, Yue; Schneiders, Eike; Naiseh, Mohammad; Fischer, Joel E.; Ramchurn, Sarvapali D.; Soorati, Mohammad D.; Archibald, Blair; Sevegnani, Michele
The effect of predictive formal modelling at runtime on performance in human-swarm interaction Proceedings Article
In: HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction, pp. 172?176, Association for Computing Machinery, 2024, (Publisher Copyright: © 2024 Copyright held by the owner/author(s)).
@inproceedings{soton488273,
title = {The effect of predictive formal modelling at runtime on performance in human-swarm interaction},
author = {Ayodeji O. Abioye and William Hunt and Yue Gu and Eike Schneiders and Mohammad Naiseh and Joel E. Fischer and Sarvapali D. Ramchurn and Mohammad D. Soorati and Blair Archibald and Michele Sevegnani},
url = {https://eprints.soton.ac.uk/488273/},
year = {2024},
date = {2024-03-01},
booktitle = {HRI '24: Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction},
pages = {172?176},
publisher = {Association for Computing Machinery},
abstract = {Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.},
note = {Publisher Copyright:
© 2024 Copyright held by the owner/author(s)},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Formal Modelling is often used as part of the design and testing process of software development to ensure that components operate within suitable bounds even in unexpected circumstances. In this paper, we use predictive formal modelling (PFM) at runtime in a human-swarm mission and show that this integration can be used to improve the performance of human-swarm teams. We recruited 60 participants to operate a simulated aerial swarm to deliver parcels to target locations. In the PFM condition, operators were informed of the estimated completion times given the number of drones deployed, whereas, in the No-PFM condition, operators did not have this information. The operators could control the mission by adding or removing drones from the mission and thereby, increasing or decreasing the overall mission cost. The evaluation of human-swarm performance relied on four metrics: the task completion time, the number of agents, the number of completed tasks, and the cost per task. Our results show that PFM modelling at runtime improves mission performance without significantly affecting the operator's workload or the system's usability.
Soorati, Mohammad D.; Naiseh, Mohammad; Hunt, William; Parnell, Katie; Clark, Jediah; Ramchurn, Sarvapali D.
Enabling trustworthiness in human-swarm systems through a digital twin Book Section
In: Dasgupta, Prithviraj; Llinas, James; Gillespie, Tony; Fouse, Scott; Lawless, William; Mittu, Ranjeev; Sofge, Donlad (Ed.): Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams, pp. 93–125, Academic Press, 2024, (Publisher Copyright: © 2024 Elsevier Inc. All rights reserved.).
@incollection{soton491769,
title = {Enabling trustworthiness in human-swarm systems through a digital twin},
author = {Mohammad D. Soorati and Mohammad Naiseh and William Hunt and Katie Parnell and Jediah Clark and Sarvapali D. Ramchurn},
editor = {Prithviraj Dasgupta and James Llinas and Tony Gillespie and Scott Fouse and William Lawless and Ranjeev Mittu and Donlad Sofge},
url = {https://eprints.soton.ac.uk/491769/},
year = {2024},
date = {2024-02-01},
booktitle = {Putting AI in the Critical Loop: Assured Trust and Autonomy in Human-Machine Teams},
pages = {93–125},
publisher = {Academic Press},
abstract = {Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.},
note = {Publisher Copyright:
© 2024 Elsevier Inc. All rights reserved.},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Robot swarms are highly dynamic systems that exhibit fault-tolerant behavior in accomplishing given tasks. Applications of swarm robotics are very limited due to the lack of complex decision-making capability. Real-world applications are only possible if we use human supervision to monitor and control the behavior of the swarm. Ensuring that human operators can trust the swarm system is one of the key challenges in human-swarm systems. This chapter presents a digital twin for trustworthy human-swarm teaming. The first element in designing such a simulation platform is to understand the trust requirements to label a human-swarm system as trustworthy. In order to outline the key trust requirements, we interviewed a group of experienced uncrewed aerial vehicle (UAV) operators and collated their suggestions for building and repairing trusts in single and multiple UAV systems. We then performed a survey to gather swarm experts? points of view on creating a taxonomy for explainability in human-swarm systems. This chapter presents a digital twin platform that implements a disaster management use case and has the capacity to meet the extracted trust and explainability requirements.
Thavanesan, Navamayooran; Parfitt, Charlotte; Bodala, Indu; Walters, Zoë; Ramchurn, Sarvapali; Underwood, Timothy; Vigneswaran, Ganesh
Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction Miscellaneous
2024.
@misc{soton497828,
title = {Machine learning models for curative and palliative oesophageal cancer treatment pathway prediction},
author = {Navamayooran Thavanesan and Charlotte Parfitt and Indu Bodala and Zoë Walters and Sarvapali Ramchurn and Timothy Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/497828/},
year = {2024},
date = {2024-01-01},
journal = {European Journal of Surgical Oncology},
volume = {50},
number = {1},
abstract = {Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Introduction: Oesophageal Cancer Multidisciplinary Teams (OC MDTs) operate under significant caseload pressures. This risks variability of decision-making which may influence patient outcomes. Machine Learning (ML) offers the ability to streamline and standardise decision-making by learning from historic treatment decisions to prediction treatment for new patients. We present internally validated ML models designed to predict OC MDT treatment decisions for curative and palliative OC patients.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Methods: four ML algorithms (multinomial logistic regression (MLR), random forests (RF), extreme gradient boost (XGB) and decision tree (DT)) were trained using nested cross-validation on a cohort of 938 OC cases from a single tertiary unit over a 12-year period. The models classified predicted treatments into one of: Surgery (S), Neoadjuvant Chemotherapy (NACT) + S, Neoadjuvant Chemoradiotherapy (NACRT) + S, Endoscopic or Palliative treatment. Performance was assessed on Area Under the Curve (AUC).ensuremath<br/ensuremath>ensuremath<br/ensuremath>Results: across algorithms, all models performed strongly with mean AUC for Surgery = 0.849$±$0.026, NACT +S = 0.884$±$0.008, NACRT +S = 0.834$±$0.035
Kelly, Thomas Graham; Soorati, Mohammad; Zauner, Klaus-Peter; Ramchurn, Gopal; Tarapore, Danesh
Trade-offs of dynamic control structure in human-swarm systems Proceedings Article
In: The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024, 2024.
@inproceedings{soton492838,
title = {Trade-offs of dynamic control structure in human-swarm systems},
author = {Thomas Graham Kelly and Mohammad Soorati and Klaus-Peter Zauner and Gopal Ramchurn and Danesh Tarapore},
url = {https://eprints.soton.ac.uk/492838/},
year = {2024},
date = {2024-01-01},
booktitle = {The International Symposium on Distributed Autonomous Robotic Systems (DARS) 2024},
abstract = {Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Swarm robotics is a study of simple robots that exhibit complex behaviour only by interacting locally with other robots and their environment. The control in swarm robotics is mainly distributed whereas centralised control is widely used in other fields of robotics. Centralised and decentralised control strategies both pose a unique set of benefits and drawbacks for the control of multi-robot systems. While decentralised systems are more scalable and resilient, they are less efficient compared to the centralised systems and they lead to excessive data transmissions to the human operators causing cognitive overload. We examine the trade-offs of each of these approaches in a human-swarm system to perform an environmental monitoring task and propose a flexible hybrid approach, which combines elements of hierarchical and decentralised systems. We find that a flexible hybrid system can outperform a centralised system (in our environmental monitoring task by 19.2%) while reducing the number of messages sent to a human operator (here by 23.1%). We conclude that establishing centralisation for a system is not always optimal for performance and that utilising aspects of centralised and decentralised systems can keep the swarm from hindering its performance.
2023
Early, Joseph; Deweese, Ying-Jung Chen; Evers, Christine; Ramchurn, Sarvapali
Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation Journal Article
In: Environmental Data Science, vol. 2, pp. 18, 2023.
@article{soton490766,
title = {Extending scene-to-patch models: Multi-resolution multiple instance learning for Earth observation},
author = {Joseph Early and Ying-Jung Chen Deweese and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/490766/},
year = {2023},
date = {2023-12-01},
journal = {Environmental Data Science},
volume = {2},
pages = {18},
abstract = {Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Land cover classification (LCC) and natural disaster response (NDR) are important issues in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation (EO) imaging data for LCC and NDR often rely on fully annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of machine learning for EO. In this study, we extend our prior work on Scene-to-Patch models: an alternative machine learning approach for EO that utilizes Multiple Instance Learning (MIL). As our approach only requires high-level scene labels, it enables much faster development of new datasets while still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using machine learning for EO. We propose new multi-resolution MIL architectures that outperform single-resolution MIL models and non-MIL baselines on the DeepGlobe LCC and FloodNet NDR datasets. In addition, we conduct a thorough analysis of model performance and interpretability.
Rigley, Eryn; Bentley, Caitlin; Krook, Joshua; Ramchurn, Gopal
Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries Journal Article
In: Global Policy, 2023, (Funding Information: This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.).
@article{soton485727,
title = {Evaluating international AI skills policy: a systematic review of AI skills policy in seven countries},
author = {Eryn Rigley and Caitlin Bentley and Joshua Krook and Gopal Ramchurn},
url = {https://eprints.soton.ac.uk/485727/},
year = {2023},
date = {2023-12-01},
journal = {Global Policy},
abstract = {ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>},
note = {Funding Information:
This research was supported via UKRI by the DCMS Science and Analysis R&D Programme. It was developed and produced according to UKRI's initial hypotheses and output requests. Any primary research, subsequent findings or recommendations do not represent Government views or policy and are produced according to academic ethics, quality assurance and independence.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>As artificial intelligence (AI) is having an increasingly disruptive impact across industries, companies continue to report having difficulty when recruiting for AI roles, while new graduates find it difficult to find employment, indicating a skills gap or skills misalignment. International approaches to AI skills programmes can offer a guide to future policy development of a skilled workforce, best placed to harness the economic opportunities that AI may support. The authors performed a systematic literature review on AI skills in government policies and documents from seven countries: Australia, Canada, China, Singapore, Sweden, the United Kingom and the United States. We found a divide between countries which emphasised a broader, nationwide approach to upskill and educate all citizens at different levels, namely the United States and Singapore and those countries which emphasised a narrower focus on educating a smaller group of experts with advanced AI knowledge and skills, namely China, Sweden and Canada. We found that the former, broader approaches tended to correlate with higher AI readiness and index scores than the narrower, expert-driven approach. Our findings indicate that, to match world-leading AI readiness, future AI skills policy should follow these broad, nationwide approaches to upskill and educate all citizens at different levels of AI expertise.ensuremath</pensuremath>
Singh, Lokesh; Ramchurn, Gopal
The effect of automated agents on individual performance under induced stress Proceedings Article
In: Kalra, Jay (Ed.): Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition), pp. 118–127, AHFE International, 2023.
@inproceedings{soton485655,
title = {The effect of automated agents on individual performance under induced stress},
author = {Lokesh Singh and Gopal Ramchurn},
editor = {Jay Kalra},
url = {https://eprints.soton.ac.uk/485655/},
year = {2023},
date = {2023-11-01},
booktitle = {Emerging Technologies in Healthcare and Medicine: Proceedings of the AHFE International Conference on Human Factors in Design, Engineering and Computing (AHFE 2023 Hawaii Edition)},
pages = {118–127},
publisher = {AHFE International},
abstract = {Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Induced stress is a phenomenon commonly experienced across different fields such as emergency services, healthcare, air traffic control, sports, and business - which necessitates the development of effective coping strategies and resilience for individuals or teams performing under pressure. This study aims to examine the effects of automated agents on individual performance during high-stress conditions. The design of these agents ensures they carry out identical tasks as participants based on predetermined frameworks. Participants underwent an experimentally designed task that aimed at inducing stress while measuring their performance amidst time pressure and auditory distraction. Results indicate that working with automated agents causes individuals to alter their approach by focusing narrowly on immediate concerns - making it challenging for them to consider several options or see broader contexts accurately. Regardless of ability level participants' performances were influenced by these automated agents. Future research will explore how these findings interact with physiological signals. This study highlights the importance of developing effective coping strategies and the potential impact of social factors on individual performance under induced stress.
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J.; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, vol. 49, no. 11, 2023, (Publisher Copyright: © 2023 The Author(s)).
@article{soton479497b,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J. Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-11-01},
journal = {European Journal of Surgical Oncology},
volume = {49},
number = {11},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
© 2023 The Author(s)},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $±$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$±$0.045] vs 0.757 [$±$0.068], 0.740 [$±$0.042], and 0.709 [$±$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub Response Miscellaneous
2023.
@misc{soton481740,
title = {AI large language models inquiry: TASHub Response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
Krook, Joshua; Williams, Jennifer; Seabrooke, Tina; Schneiders, Eike; Blockx, Jan; Middleton, Stuart E; Ramchurn, Sarvapali
AI large language models inquiry: TASHub response Miscellaneous
2023.
@misc{soton481740b,
title = {AI large language models inquiry: TASHub response},
author = {Joshua Krook and Jennifer Williams and Tina Seabrooke and Eike Schneiders and Jan Blockx and Stuart E Middleton and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/481740/},
year = {2023},
date = {2023-08-01},
publisher = {University of Southampton},
abstract = {Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Policy submission to the Consultation by Communications and Digital Committee, House of Lords, AI Large Language Models Inquiry.ensuremath<br/ensuremath>
Thavanesan, Navamayooran; Bodala, Indu; Walters, Zoe; Ramchurn, Sarvapali; Underwood, Timothy J; Vigneswaran, Ganesh
Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer Journal Article
In: European Journal of Surgical Oncology, 2023, (Publisher Copyright: copyright 2023 The Author(s)).
@article{soton479497,
title = {Machine learning to predict curative multidisciplinary team treatment decisions in oesophageal cancer},
author = {Navamayooran Thavanesan and Indu Bodala and Zoe Walters and Sarvapali Ramchurn and Timothy J Underwood and Ganesh Vigneswaran},
url = {https://eprints.soton.ac.uk/479497/},
year = {2023},
date = {2023-07-01},
journal = {European Journal of Surgical Oncology},
abstract = {ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>},
note = {Publisher Copyright:
copyright 2023 The Author(s)},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>Background: Rising workflow pressures within the oesophageal cancer (OC) multidisciplinary team (MDT) can lead to variability in decision-making, and health inequality. Machine learning (ML) offers a potential automated data-driven approach to address inconsistency and standardize care. The aim of this experimental pilot study was to develop ML models able to predict curative OC MDT treatment decisions and determine the relative importance of underlying decision-critical variables. Methods: Retrospective complete-case analysis of oesophagectomy patients $pm$ neoadjuvant chemotherapy (NACT) or chemoradiotherapy (NACRT) between 2010 and 2020. Established ML algorithms (Multinomial Logistic regression (MLR), Random Forests (RF), Extreme Gradient Boosting (XGB)) and Decision Tree (DT) were used to train models predicting OC MDT treatment decisions: surgery (S), NACT + S or NACRT + S. Performance metrics included Area Under the Curve (AUC), Accuracy, Kappa, LogLoss, F1 and Precision -Recall AUC. Variable importance was calculated for each model. Results: We identified 399 cases with a male-to-female ratio of 3.6:1 and median age of 66.1yrs (range 32?83). MLR outperformed RF, XGB and DT across performance metrics (mean AUC of 0.793 [$pm$0.045] vs 0.757 [$pm$0.068], 0.740 [$pm$0.042], and 0.709 [$pm$0.021] respectively). Variable importance analysis identified age as a major factor in the decision to offer surgery alone or NACT + S across models (p < 0.05). Conclusions: ML techniques can use limited feature-sets to predict curative UGI MDT treatment decisions. Explainable Artificial Intelligence methods provide insight into decision-critical variables, highlighting underlying subconscious biases in cancer care decision-making. Such models may allow prioritization of caseload, improve efficiency, and offer data-driven decision-assistance to MDTs in the future.ensuremath</pensuremath>
Abioye, Ayodeji
University of Southampton, 2023.
@phdthesis{soton479472,
title = {Multimodal speech and visual gesture control interface technique for small unmanned multirotor aircraft},
author = {Ayodeji Abioye},
url = {https://eprints.soton.ac.uk/479472/},
year = {2023},
date = {2023-07-01},
publisher = {University of Southampton},
school = {University of Southampton},
abstract = {ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
ensuremath<p class="MsoNormal"ensuremath>This research conducted an investigation into the use of novel human computer interaction(HCI) interfaces in the control of small multirotor unmanned aerial vehicles(UAVs). The main objective was to propose, design, and develop an alternative control interface for the small multirotor UAV, which could perform better than the standard RC joystick (RCJ) controller, and to evaluate the performance of the proposed interface. The multimodal speech and visual gesture (mSVG)interface were proposed, designed, and developed. This was then coupled to a Rotor S ROS Gazebo UAV simulator. An experiment study was designed to determine how practical the use of the proposed multimodal speech and visual gesture interface was in the control of small multirotor UAVs by determining the limits of speech and gesture at different ambient noise levels and under different background-lighting conditions, respectively. And to determine how the mSVG interface compares to the RC joystick controller for a simple navigational control task - in terms of performance (time of completion and accuracy of navigational control) and from a human factor?s perspective (user satisfaction and cognitive workload). 37 participants were recruited. From the results of the experiments conducted, the mSVG interface was found to be an effective alternative to the RCJ interface when operated within a constrained application environment. From the result of the noise level experiment, it was observed that speech recognition accuracy/success rate falls as noise levels rise, with75 dB noise level being the practical aerial robot (aerobot) application limit. From the results of the gesture lighting experiment, gestures were successfully recognised from 10 Lux and above on distinct solid backgrounds, but the effect of varying both the lighting conditions and the environment background on the quality of gesture recognition, was insignificant (< 0.5%), implying that the technology used, type of gesture captured, and the image processing technique used were more important. From the result of the performance and cognitive workload comparison between the RCJ and mSVG interfaces, the mSVG interface was found to perform better at higher nCA application levels than the RCJ interface. The mSVG interface was 1 minute faster and 25% more accurate than the RCJ interface; and the RCJ interface was found to be 1.4 times more cognitively demanding than the mSVG interface. The main limitation of this research was the limited lighting level range of 10 Lux - 1400 Lux used during the gesture lighting experiment, which constrains the application limit to lowlighting indoor environments. Suggested further works from this research included the development of a more robust gesture and speech algorithm and the coupling of the improved mSVG interface on to a practical UAV.ensuremath</pensuremath>
Abioye, Ayodeji; Naiseh, Mohammad; Hunt, William; Clark, Jediah R; Ramchurn, Sarvapali D; Soorati, Mohammad
The effect of data visualisation quality and task density on human-swarm interaction Proceedings Article
In: Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), IEEE, 2023.
@inproceedings{soton479970,
title = {The effect of data visualisation quality and task density on human-swarm interaction},
author = {Ayodeji Abioye and Mohammad Naiseh and William Hunt and Jediah R Clark and Sarvapali D Ramchurn and Mohammad Soorati},
url = {https://eprints.soton.ac.uk/479970/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
booktitle = {Proceedings of the 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
publisher = {IEEE},
abstract = {Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Despite the advantages of having robot swarms, human supervision is required for real-world applications. The performance of the human-swarm system depends on several factors including the data availability for the human operators. In this paper, we study the human factors aspect of the human-swarm interaction and investigate how having access to high-quality data can affect the performance of the human-swarm system - the number of tasks completed and the human trust level in operation. We designed an experiment where a human operator is tasked to operate a swarm to identify casualties in an area within a given time period. One group of operators had the option to request high-quality pictures while the other group had to base their decision on the available low-quality images. We performed a user study with 120 participants and recorded their success rate (directly logged via the simulation platform) as well as their workload and trust level (measured through a questionnaire after completing a human-swarm scenario). The findings from our study indicated that the group granted access to high-quality data exhibited an increased workload and placed greater trust in the swarm, thus confirming our initial hypothesis. However, we also found that the number of accurately identified casualties did not significantly vary between the two groups, suggesting that data quality had no impact on the successful completion of tasks.
Krook, Joshua; McAuley, Derek; Anderson, Stuart; Downer, John; Winter, Peter; Ramchurn, Sarvapali D
AI Foundation Models: initial review, CMA Consultation, TAS Hub Response Miscellaneous
2023.
@misc{soton477553,
title = {AI Foundation Models: initial review, CMA Consultation, TAS Hub Response},
author = {Joshua Krook and Derek McAuley and Stuart Anderson and John Downer and Peter Winter and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477553/},
year = {2023},
date = {2023-06-01},
urldate = {2023-06-01},
publisher = {University of Southampton},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Krook, Joshua; Downer, John; Winter, Peter; Williams, Jennifer; Ives, Jonathan; Bratu, Roxana; Sheir, Stephanie; Williams, Robin; Anderson, Stuart; Li, Phoebe; Ramamoorthy, Subramanian; Ramchurn, Sarvapali
AI regulation: a pro-innovation approach ? policy proposals: TASHub Response Miscellaneous
2023.
@misc{soton478329,
title = {AI regulation: a pro-innovation approach ? policy proposals: TASHub Response},
author = {Joshua Krook and John Downer and Peter Winter and Jennifer Williams and Jonathan Ives and Roxana Bratu and Stephanie Sheir and Robin Williams and Stuart Anderson and Phoebe Li and Subramanian Ramamoorthy and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478329/},
year = {2023},
date = {2023-06-01},
publisher = {University of Southampton},
abstract = {Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Response to open consultation from: Department for Science, Innovation and Technologyensuremath<br/ensuremath>and Office for Artificial Intelligence
Hunt, William; Ryan, Jack; Abioye, Ayodeji O; Ramchurn, Sarvapali D; Soorati, Mohammad D
Demonstrating performance benefits of human-swarm teaming Proceedings Article
In: Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems, pp. 3062–3064, International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS), 2023.
@inproceedings{soton479903,
title = {Demonstrating performance benefits of human-swarm teaming},
author = {William Hunt and Jack Ryan and Ayodeji O Abioye and Sarvapali D Ramchurn and Mohammad D Soorati},
url = {https://eprints.soton.ac.uk/479903/},
year = {2023},
date = {2023-05-01},
urldate = {2023-05-01},
booktitle = {Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems},
pages = {3062–3064},
publisher = {International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS)},
abstract = {Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Autonomous swarms of robots can bring robustness, scalability and adaptability to safety-critical tasks such as search and rescue but their application is still very limited. Using semi-autonomous swarms with human control can bring robot swarms to real-world applications. Human operators can define goals for the swarm, monitor their performance and interfere with, or overrule, the decisions and behaviour. We present the "Human And Robot Interactive Swarm'' simulator (HARIS) that allows multi-user interaction with a robot swarm and facilitates qualitative and quantitative user studies through simulation of robot swarms completing tasks, from package delivery to search and rescue, with varying levels of human control. In this demonstration, we showcase the simulator by using it to study the performance gain offered by maintaining a "human-in-the-loop'' over a fully autonomous system as an example. This is illustrated in the context of search and rescue, with an autonomous allocation of resources to those in need.
Worrawichaipat, Phuriwat; Gerding, Enrico; Kaparias, Ioannis; Ramchurn, Sarvapali
Multi-agent signal-less intersection management with dynamic platoon formation Proceedings Article
In: 22nd International Conference on Autonomous Agents and Multiagent Systems (29/05/23 - 02/06/23), pp. 1542–1550, 2023.
@inproceedings{soton478647,
title = {Multi-agent signal-less intersection management with dynamic platoon formation},
author = {Phuriwat Worrawichaipat and Enrico Gerding and Ioannis Kaparias and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/478647/},
year = {2023},
date = {2023-05-01},
urldate = {2023-05-01},
booktitle = {22nd International Conference on Autonomous Agents and Multiagent Systems (29/05/23 - 02/06/23)},
pages = {1542–1550},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Everett, Gregory; Beal, Ryan J; Matthews, Tim; Early, Joseph; Norman, Timothy J; Ramchurn, Sarvapali D
Inferring player location in sports matches: multi-agent spatial imputation from limited observations Miscellaneous
2023.
@misc{soton477020,
title = {Inferring player location in sports matches: multi-agent spatial imputation from limited observations},
author = {Gregory Everett and Ryan J Beal and Tim Matthews and Joseph Early and Timothy J Norman and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/477020/},
year = {2023},
date = {2023-02-01},
abstract = {Understanding agent behaviour in Multi-Agent Systems (MAS) is an important problem in domains such as autonomous driving, disaster response, and sports analytics. Existing MAS problems typically use uniform timesteps with observations for all agents. In this work, we analyse the problem of agent location imputation, specifically posed in environments with non-uniform timesteps and limited agent observability (textttchar12695% missing values). Our approach uses Long Short-Term Memory and Graph Neural Network components to learn temporal and inter-agent patterns to predict the location of all agents at every timestep. We apply this to the domain of football (soccer) by imputing the location of all players in a game from sparse event data (e.g., shots and passes). Our model estimates player locations to within textttchar1266.9m; a textttchar12662% reduction in error from the best performing baseline. This approach facilitates downstream analysis tasks such as player physical metrics, player coverage, and team pitch control. Existing solutions to these tasks often require optical tracking data, which is expensive to obtain and only available to elite clubs. By imputing player locations from easy to obtain event data, we increase the accessibility of downstream tasks.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Understanding agent behaviour in Multi-Agent Systems (MAS) is an important problem in domains such as autonomous driving, disaster response, and sports analytics. Existing MAS problems typically use uniform timesteps with observations for all agents. In this work, we analyse the problem of agent location imputation, specifically posed in environments with non-uniform timesteps and limited agent observability (textttchar12695% missing values). Our approach uses Long Short-Term Memory and Graph Neural Network components to learn temporal and inter-agent patterns to predict the location of all agents at every timestep. We apply this to the domain of football (soccer) by imputing the location of all players in a game from sparse event data (e.g., shots and passes). Our model estimates player locations to within textttchar1266.9m; a textttchar12662% reduction in error from the best performing baseline. This approach facilitates downstream analysis tasks such as player physical metrics, player coverage, and team pitch control. Existing solutions to these tasks often require optical tracking data, which is expensive to obtain and only available to elite clubs. By imputing player locations from easy to obtain event data, we increase the accessibility of downstream tasks.
2022
Ahmed, Sarah; Azim, Tayyaba; Early, Joseph Arthur; Ramchurn, Sarvapali
Revisiting deep fisher vectors: using fisher information to improve object classification Proceedings Article
In: British Machine Vision Conference (21/11/22 - 24/11/22), 2022.
@inproceedings{soton471260,
title = {Revisiting deep fisher vectors: using fisher information to improve object classification},
author = {Sarah Ahmed and Tayyaba Azim and Joseph Arthur Early and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/471260/},
year = {2022},
date = {2022-11-01},
booktitle = {British Machine Vision Conference (21/11/22 - 24/11/22)},
abstract = {Although deep learning models have become the gold standard in achieving outstanding results on a large variety of computer vision and machine learning tasks, the use of kernel methods has still not gone out of trend because of its potential to beat deep learning performances at a number of occasions. Given the potential of kernel techniques, prior works have also proposed the use of hybrid approaches combining deep learning with kernel learning to complement their respective strengths and weaknesses. This work develops this idea further by introducing an improved version of Fisher kernels derived from the deep Boltzmann machines (DBM). Our improved deep Fisher kernel (IDFK) utilises an approximation of the Fisher information matrix to derive improved Fisher vectors. We show IDFK can be utilised to retain a high degree of class separability, making it appropriate for classification and retrieval tasks. The efficacy of the proposed approach is evaluated on three benchmark data sets: MNIST, USPS and Alphanumeric, showing an improvement in classification performance over existing kernel approaches, and comparable performance to deep learning methods, but with much reduced computational costs. Using explainable AI methods, we also demonstrate why our IDFK leads to better classification performance.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Although deep learning models have become the gold standard in achieving outstanding results on a large variety of computer vision and machine learning tasks, the use of kernel methods has still not gone out of trend because of its potential to beat deep learning performances at a number of occasions. Given the potential of kernel techniques, prior works have also proposed the use of hybrid approaches combining deep learning with kernel learning to complement their respective strengths and weaknesses. This work develops this idea further by introducing an improved version of Fisher kernels derived from the deep Boltzmann machines (DBM). Our improved deep Fisher kernel (IDFK) utilises an approximation of the Fisher information matrix to derive improved Fisher vectors. We show IDFK can be utilised to retain a high degree of class separability, making it appropriate for classification and retrieval tasks. The efficacy of the proposed approach is evaluated on three benchmark data sets: MNIST, USPS and Alphanumeric, showing an improvement in classification performance over existing kernel approaches, and comparable performance to deep learning methods, but with much reduced computational costs. Using explainable AI methods, we also demonstrate why our IDFK leads to better classification performance.
Yazdanpanah, Vahid; Gerding, Enrico; Stein, Sebastian; Dastani, Mehdi; Jonker, Catholijn M; Norman, Timothy; Ramchurn, Sarvapali
Reasoning About Responsibility in Autonomous Systems: Challenges and Opportunities Journal Article
In: AI & Society, 2022.
@article{soton471971,
title = {Reasoning About Responsibility in Autonomous Systems: Challenges and Opportunities},
author = {Vahid Yazdanpanah and Enrico Gerding and Sebastian Stein and Mehdi Dastani and Catholijn M Jonker and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/471971/},
year = {2022},
date = {2022-11-01},
journal = {AI & Society},
abstract = {Ensuring the trustworthiness of autonomous systems and artificial intelligenceensuremath<br/ensuremath>is an important interdisciplinary endeavour. In this position paper, we argue thatensuremath<br/ensuremath>this endeavour will benefit from technical advancements in capturing various forms of responsibility, and we present a comprehensive research agenda to achieve this. In particular, we argue that ensuring the reliability of autonomous system can take advantage of technical approaches for quantifying degrees of responsibility and for coordinating tasks based on that. Moreover, we deem that, in certifying the legality of an AI system, formal and computationally implementable notions of responsibility, blame, accountability, and liability are applicable for addressing potential responsibility gaps (i.e., situations in which a group is responsible, but individuals? responsibility may be unclear). This is a call to enable AI systems themselves, as well as those involved in the design, monitoring, and governance of AI systems, to represent and reason about who can be seen as responsible in prospect (e.g., for completing a task in future) and who can be seen as responsible retrospectively (e.g., for a failure that has already occurred). To that end, in this work, we show that across all stages of the design, development, and deployment of Trustworthy Autonomous Systems (TAS), responsibility reasoning should play a key role. This position paper is the first step towards establishing a road-map and research agenda on how the notion of responsibility can provide novel solution concepts for ensuring the reliability and legality of TAS and, as a result, enables an effective embedding of AI technologies into society.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Ensuring the trustworthiness of autonomous systems and artificial intelligenceensuremath<br/ensuremath>is an important interdisciplinary endeavour. In this position paper, we argue thatensuremath<br/ensuremath>this endeavour will benefit from technical advancements in capturing various forms of responsibility, and we present a comprehensive research agenda to achieve this. In particular, we argue that ensuring the reliability of autonomous system can take advantage of technical approaches for quantifying degrees of responsibility and for coordinating tasks based on that. Moreover, we deem that, in certifying the legality of an AI system, formal and computationally implementable notions of responsibility, blame, accountability, and liability are applicable for addressing potential responsibility gaps (i.e., situations in which a group is responsible, but individuals? responsibility may be unclear). This is a call to enable AI systems themselves, as well as those involved in the design, monitoring, and governance of AI systems, to represent and reason about who can be seen as responsible in prospect (e.g., for completing a task in future) and who can be seen as responsible retrospectively (e.g., for a failure that has already occurred). To that end, in this work, we show that across all stages of the design, development, and deployment of Trustworthy Autonomous Systems (TAS), responsibility reasoning should play a key role. This position paper is the first step towards establishing a road-map and research agenda on how the notion of responsibility can provide novel solution concepts for ensuring the reliability and legality of TAS and, as a result, enables an effective embedding of AI technologies into society.
Early, Joseph; Deweese, Ying-Jung; Evers, Christine; Ramchurn, Sarvapali
Scene-to-Patch earth observation: multiple instance learning for land cover classification Miscellaneous
2022, (14 pages total (4 main content; 2 acknowledgments + citations; 8 appendices); 8 figures (2 main; 6 appendix); published at "Tackling Climate Change with Machine Learning: Workshop at NeurIPS 2022").
@misc{soton472853,
title = {Scene-to-Patch earth observation: multiple instance learning for land cover classification},
author = {Joseph Early and Ying-Jung Deweese and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/472853/},
year = {2022},
date = {2022-11-01},
abstract = {Land cover classification (LCC), and monitoring how land use changes over time, is an important process in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation data for LCC rely on fully-annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of LCC. In this study, we propose Scene-to-Patch models: an alternative LCC approach utilising Multiple Instance Learning (MIL) that requires only high-level scene labels. This enables much faster development of new datasets whilst still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using LCC for different scenarios. On the DeepGlobe-LCC dataset, our approach outperforms non-MIL baselines on both scene- and patch-level prediction. This work provides the foundation for expanding the use of LCC in climate change mitigation methods for technology, government, and academia.},
note = {14 pages total (4 main content; 2 acknowledgments + citations; 8 appendices); 8 figures (2 main; 6 appendix); published at "Tackling Climate Change with Machine Learning: Workshop at NeurIPS 2022"},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Land cover classification (LCC), and monitoring how land use changes over time, is an important process in climate change mitigation and adaptation. Existing approaches that use machine learning with Earth observation data for LCC rely on fully-annotated and segmented datasets. Creating these datasets requires a large amount of effort, and a lack of suitable datasets has become an obstacle in scaling the use of LCC. In this study, we propose Scene-to-Patch models: an alternative LCC approach utilising Multiple Instance Learning (MIL) that requires only high-level scene labels. This enables much faster development of new datasets whilst still providing segmentation through patch-level predictions, ultimately increasing the accessibility of using LCC for different scenarios. On the DeepGlobe-LCC dataset, our approach outperforms non-MIL baselines on both scene- and patch-level prediction. This work provides the foundation for expanding the use of LCC in climate change mitigation methods for technology, government, and academia.
Parnell, Katie; Fischer, Joel E; Clark, Jediah R; Bodenmann, Adrian; Trigo, Maria Jose Galvez; Brito, Mario; Soorati, Mohammad Divband; Plant, Katherine; Ramchurn, Sarvapali
Trustworthy UAV relationships: Applying the Schema Action World taxonomy to UAVs and UAV swarm operations Journal Article
In: International Journal of Human-Computer Interaction, 2022.
@article{soton468839,
title = {Trustworthy UAV relationships: Applying the Schema Action World taxonomy to UAVs and UAV swarm operations},
author = {Katie Parnell and Joel E Fischer and Jediah R Clark and Adrian Bodenmann and Maria Jose Galvez Trigo and Mario Brito and Mohammad Divband Soorati and Katherine Plant and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/468839/},
year = {2022},
date = {2022-07-01},
journal = {International Journal of Human-Computer Interaction},
abstract = {Human Factors play a significant role inthe development and integration of avionic systems to ensure that they are trusted and can be used effectively. As Unoccupied Aerial Vehicle (UAV) technology becomes increasingly important to the aviation domain this holds true. This study aims to gain an understanding of UAV operators?trust requirements when piloting UAVs by utilising a popular aviation interview methodology (Schema World Action Research Method), in combination with key questions on trust identified from the literature. Interviews were conducted with six UAVoperators, with a range of experience. This identified the importance of past experience to trust and the expectations that operators hold. Recommendations are made that target training to inform experience, in addition to the equipment, procedures and organisational standards that can aid in developing trustworthy systems. The methodology that was developed shows promise for capturing trust within human-automation interactions},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Human Factors play a significant role inthe development and integration of avionic systems to ensure that they are trusted and can be used effectively. As Unoccupied Aerial Vehicle (UAV) technology becomes increasingly important to the aviation domain this holds true. This study aims to gain an understanding of UAV operators?trust requirements when piloting UAVs by utilising a popular aviation interview methodology (Schema World Action Research Method), in combination with key questions on trust identified from the literature. Interviews were conducted with six UAVoperators, with a range of experience. This identified the importance of past experience to trust and the expectations that operators hold. Recommendations are made that target training to inform experience, in addition to the equipment, procedures and organisational standards that can aid in developing trustworthy systems. The methodology that was developed shows promise for capturing trust within human-automation interactions
Soorati, Mohammad Divband; Gerding, Enrico; Marchioni, Enrico; Naumov, Pavel; Norman, Timothy; Ramchurn, Sarvapali; Rastegari, Baharak; Sobey, Adam; Stein, Sebastian; Tarapore, Danesh; Yazdanpanah, Vahid; Zhang, Jie
From Intelligent Agents to Trustworthy Human-Centred Multiagent Systems Journal Article
In: AI Communications, 2022.
@article{soton467975,
title = {From Intelligent Agents to Trustworthy Human-Centred Multiagent Systems},
author = {Mohammad Divband Soorati and Enrico Gerding and Enrico Marchioni and Pavel Naumov and Timothy Norman and Sarvapali Ramchurn and Baharak Rastegari and Adam Sobey and Sebastian Stein and Danesh Tarapore and Vahid Yazdanpanah and Jie Zhang},
url = {https://eprints.soton.ac.uk/467975/},
year = {2022},
date = {2022-07-01},
journal = {AI Communications},
abstract = {The Agents, Interaction and Complexity research group at the University of Southampton has a long track record of research in multiagent systems (MAS). We have made substantial scientific contributions across learning in MAS, game-theoretic techniques for coordinating agent systems, and formal methods for representation and reasoning. We highlight key results achieved by the group and elaborate on recent work and open research challenges in developing trustworthy autonomous systems and deploying human-centred AI systems that aim to support societal good.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The Agents, Interaction and Complexity research group at the University of Southampton has a long track record of research in multiagent systems (MAS). We have made substantial scientific contributions across learning in MAS, game-theoretic techniques for coordinating agent systems, and formal methods for representation and reasoning. We highlight key results achieved by the group and elaborate on recent work and open research challenges in developing trustworthy autonomous systems and deploying human-centred AI systems that aim to support societal good.
Bossens, David; Ramchurn, Sarvapali; Tarapore, Danesh
Resilient robot teams: a review integrating decentralised control, change-detection, and learning Miscellaneous
2022.
@misc{soton457101,
title = {Resilient robot teams: a review integrating decentralised control, change-detection, and learning},
author = {David Bossens and Sarvapali Ramchurn and Danesh Tarapore},
url = {https://eprints.soton.ac.uk/457101/},
year = {2022},
date = {2022-06-01},
journal = {Current Robotics Reports},
abstract = {Purpose of review: This paper reviews opportunities and challenges for decentralised control, change-detection, and learning in the context of resilient robot teams.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Recent findings: Exogenous fault detection methods can provide a generic detection or a specific diagnosis with a recovery solution. Robot teams can perform active and distributed sensing for detecting changes in the environment, including identifying and tracking dynamic anomalies, as well as collaboratively mapping dynamic environments. Resilient methods for decentralised control have been developed in learning perception-action-communication loops, multi-agent reinforcement learning, embodied evolution, offline evolution with online adaptation, explicit task allocation, and stigmergy in swarm robotics.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Summary: Remaining challenges for resilient robot teams are integrating change-detection and trial-and-error learning methods, obtaining reliable performance evaluations under constrained evaluation time, improving the safety of resilient robot teams, theoretical results demonstrating rapid adaptation to given environmental perturbations, and designing realistic and compelling case studies.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Purpose of review: This paper reviews opportunities and challenges for decentralised control, change-detection, and learning in the context of resilient robot teams.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Recent findings: Exogenous fault detection methods can provide a generic detection or a specific diagnosis with a recovery solution. Robot teams can perform active and distributed sensing for detecting changes in the environment, including identifying and tracking dynamic anomalies, as well as collaboratively mapping dynamic environments. Resilient methods for decentralised control have been developed in learning perception-action-communication loops, multi-agent reinforcement learning, embodied evolution, offline evolution with online adaptation, explicit task allocation, and stigmergy in swarm robotics.ensuremath<br/ensuremath>ensuremath<br/ensuremath>Summary: Remaining challenges for resilient robot teams are integrating change-detection and trial-and-error learning methods, obtaining reliable performance evaluations under constrained evaluation time, improving the safety of resilient robot teams, theoretical results demonstrating rapid adaptation to given environmental perturbations, and designing realistic and compelling case studies.
Early, Joseph; Bewley, Tom; Evers, Christine; Ramchurn, Sarvapali
Non-markovian reward modelling from trajectory labels via interpretable multiple instance learning Journal Article
In: arXiv, 2022, (20 pages (9 main content; 2 references; 9 appendix). 11 figures (8 main content; 3 appendix)).
@article{soton458023,
title = {Non-markovian reward modelling from trajectory labels via interpretable multiple instance learning},
author = {Joseph Early and Tom Bewley and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/458023/},
year = {2022},
date = {2022-05-01},
journal = {arXiv},
abstract = {We generalise the problem of reward modelling (RM) for reinforcement learning (RL) to handle non-Markovian rewards. Existing work assumes that human evaluators observe each step in a trajectory independently when providing feedback on agent behaviour. In this work, we remove this assumption, extending RM to include hidden state information that captures temporal dependencies in human assessment of trajectories. We then show how RM can be approached as a multiple instance learning (MIL) problem, and develop new MIL models that are able to capture the time dependencies in labelled trajectories. We demonstrate on a range of RL tasks that our novel MIL models can reconstruct reward functions to a high level of accuracy, and that they provide interpretable learnt hidden information that can be used to train high-performing agent policies.},
note = {20 pages (9 main content; 2 references; 9 appendix). 11 figures (8 main content; 3 appendix)},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
We generalise the problem of reward modelling (RM) for reinforcement learning (RL) to handle non-Markovian rewards. Existing work assumes that human evaluators observe each step in a trajectory independently when providing feedback on agent behaviour. In this work, we remove this assumption, extending RM to include hidden state information that captures temporal dependencies in human assessment of trajectories. We then show how RM can be approached as a multiple instance learning (MIL) problem, and develop new MIL models that are able to capture the time dependencies in labelled trajectories. We demonstrate on a range of RL tasks that our novel MIL models can reconstruct reward functions to a high level of accuracy, and that they provide interpretable learnt hidden information that can be used to train high-performing agent policies.
Buermann, Jan; Georgiev, Dimitar; Gerding, Enrico; Hill, Lewis; Malik, Obaid; Pop, Alexandru; Pun, Matthew; Ramchurn, Sarvapali; Salisbury, Elliot; Stojanovic, Ivan
An agent-based simulator for maritime transport decarbonisation: Demonstration track Proceedings Article
In: 21st International Conference on Autonomous Agents and Multiagent Systems (09/05/22 - 13/05/22), pp. 1890–1892, 2022.
@inproceedings{soton456716,
title = {An agent-based simulator for maritime transport decarbonisation: Demonstration track},
author = {Jan Buermann and Dimitar Georgiev and Enrico Gerding and Lewis Hill and Obaid Malik and Alexandru Pop and Matthew Pun and Sarvapali Ramchurn and Elliot Salisbury and Ivan Stojanovic},
url = {https://eprints.soton.ac.uk/456716/},
year = {2022},
date = {2022-05-01},
booktitle = {21st International Conference on Autonomous Agents and Multiagent Systems (09/05/22 - 13/05/22)},
pages = {1890–1892},
abstract = {Greenhouse gas (GHG) emission reduction is an important and necessary goal; currently, different policies to reduce GHG emissions in maritime transport are being discussed. Amongst policies, like carbon taxes or carbon intensity targets, it is hard to determine which policies can successfully reduce GHG emissions while allowing the industry to be profitable. We introduce an agent-based maritime transport simulator to investigate the effectiveness of two decarbonisation policies by simulating a maritime transport operator?s trade pattern and fleet make-up changes as a reaction to taxation and fixed targets. This first of its kind simulator allows to compare and quantify the difference of carbon reduction policies and how they affect shipping operations.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Greenhouse gas (GHG) emission reduction is an important and necessary goal; currently, different policies to reduce GHG emissions in maritime transport are being discussed. Amongst policies, like carbon taxes or carbon intensity targets, it is hard to determine which policies can successfully reduce GHG emissions while allowing the industry to be profitable. We introduce an agent-based maritime transport simulator to investigate the effectiveness of two decarbonisation policies by simulating a maritime transport operator?s trade pattern and fleet make-up changes as a reaction to taxation and fixed targets. This first of its kind simulator allows to compare and quantify the difference of carbon reduction policies and how they affect shipping operations.
Rigas, Emmanouil S; Gerding, Enrico H; Stein, Sebastian; Ramchurn, Sarvapali D; Bassiliades, Nick
Mechanism design for efficient offline and online allocation of electric vehicles to charging stations Journal Article
In: Energies, vol. 15, no. 5, 2022, (Funding Information: Funding: This research study was co-financed by Greece and the European Union (European Social Fund–ESF) through the Operational Programme ?Human Resources Development, Education and Lifelong Learning? in the context of the project ?Reinforcement of Postdoctoral Researchers-2nd Cycle? (MIS-5033021), implemented by State Scholarships Foundation (IKY). Copyright 2022 Elsevier B.V., All rights reserved.).
@article{soton455806,
title = {Mechanism design for efficient offline and online allocation of electric vehicles to charging stations},
author = {Emmanouil S Rigas and Enrico H Gerding and Sebastian Stein and Sarvapali D Ramchurn and Nick Bassiliades},
url = {https://eprints.soton.ac.uk/455806/},
year = {2022},
date = {2022-03-01},
journal = {Energies},
volume = {15},
number = {5},
abstract = {ensuremath<pensuremath>The industry related to electric vehicles (EVs) has seen a substantial increase in recent years, as such vehicles have the ability to significantly reduce total COensuremath<subensuremath>2ensuremath</subensuremath> emissions and the related global warming effect. In this paper, we focus on the problem of allocating EVs to charging stations, scheduling and pricing their charging. Specifically, we developed a Mixed Integer Program (MIP) which executes offline and optimally allocates EVs to charging stations. On top, we propose two alternative mechanisms to price the electricity the EVs charge. The first mechanism is a typical fixed-price one, while the second is a variation of the Vickrey?Clark?Groves (VCG) mechanism. We also developed online solutions that incrementally call the MIP-based algorithm and solve it for branches of EVs. In all cases, the EVs? aim is to minimize the price to pay and the impact on their driving schedule, acting as self-interested agents. We conducted a thorough empirical evaluation of our mechanisms and we observed that they had satisfactory scalability. Additionally, the VCG mechanism achieved an up to 2.2% improvement in terms of the number of vehicles that were charged compared to the fixed-price one and, in cases where the stations were congested, it calculated higher prices for the EVs and provided a higher profit for the stations, but lower utility to the EVs. However, in a theoretical evaluation, we proved that the variant of the VCG mechanism being proposed in this paper still guaranteed truthful reporting of the EVs? preferences. In contrast, the fixed-price one was found to be vulnerable to agents? strategic behavior as non-truthful EVs can charge instead of truthful ones. Finally, we observed the online algorithms to be, on average, at 95.6% of the offline ones in terms of the average number of serviced EVs.ensuremath</pensuremath>},
note = {Funding Information:
Funding: This research study was co-financed by Greece and the European Union (European Social Fund–ESF) through the Operational Programme ?Human Resources Development, Education and Lifelong Learning? in the context of the project ?Reinforcement of Postdoctoral Researchers-2nd Cycle? (MIS-5033021), implemented by State Scholarships Foundation (IKY).
Copyright 2022 Elsevier B.V., All rights reserved.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
ensuremath<pensuremath>The industry related to electric vehicles (EVs) has seen a substantial increase in recent years, as such vehicles have the ability to significantly reduce total COensuremath<subensuremath>2ensuremath</subensuremath> emissions and the related global warming effect. In this paper, we focus on the problem of allocating EVs to charging stations, scheduling and pricing their charging. Specifically, we developed a Mixed Integer Program (MIP) which executes offline and optimally allocates EVs to charging stations. On top, we propose two alternative mechanisms to price the electricity the EVs charge. The first mechanism is a typical fixed-price one, while the second is a variation of the Vickrey?Clark?Groves (VCG) mechanism. We also developed online solutions that incrementally call the MIP-based algorithm and solve it for branches of EVs. In all cases, the EVs? aim is to minimize the price to pay and the impact on their driving schedule, acting as self-interested agents. We conducted a thorough empirical evaluation of our mechanisms and we observed that they had satisfactory scalability. Additionally, the VCG mechanism achieved an up to 2.2% improvement in terms of the number of vehicles that were charged compared to the fixed-price one and, in cases where the stations were congested, it calculated higher prices for the EVs and provided a higher profit for the stations, but lower utility to the EVs. However, in a theoretical evaluation, we proved that the variant of the VCG mechanism being proposed in this paper still guaranteed truthful reporting of the EVs? preferences. In contrast, the fixed-price one was found to be vulnerable to agents? strategic behavior as non-truthful EVs can charge instead of truthful ones. Finally, we observed the online algorithms to be, on average, at 95.6% of the offline ones in terms of the average number of serviced EVs.ensuremath</pensuremath>
Early, Joseph; Evers, Christine; Ramchurn, Sarvapali
Model agnostic interpretability for multiple instance learning Proceedings Article
In: International Conference on Learning Representations 2022 (25/04/22 - 29/04/22), 2022, (25 pages (9 content, 2 acknowledgement + references, 14 appendix). 16 figures (3 main content, 13 appendix). Submitted and accepted to ICLR 22, see http://openreview.net/forum?id=KSSfF5lMIAg . Revision: added additional acknowledgements).
@inproceedings{soton454952,
title = {Model agnostic interpretability for multiple instance learning},
author = {Joseph Early and Christine Evers and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/454952/},
year = {2022},
date = {2022-01-01},
booktitle = {International Conference on Learning Representations 2022 (25/04/22 - 29/04/22)},
abstract = {In Multiple Instance Learning (MIL), models are trained using bags of instances, where only a single label is provided for each bag. A bag label is often only determined by a handful of key instances within a bag, making it difficult to interpret what information a classifier is using to make decisions. In this work, we establish the key requirements for interpreting MIL models. We then go on to develop several model-agnostic approaches that meet these requirements. Our methods are compared against existing inherently interpretable MIL models on several datasets, and achieve an increase in interpretability accuracy of up to 30%. We also examine the ability of the methods to identify interactions between instances and scale to larger datasets, improving their applicability to real-world problems.},
note = {25 pages (9 content, 2 acknowledgement + references, 14 appendix). 16 figures (3 main content, 13 appendix). Submitted and accepted to ICLR 22, see http://openreview.net/forum?id=KSSfF5lMIAg . Revision: added additional acknowledgements},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In Multiple Instance Learning (MIL), models are trained using bags of instances, where only a single label is provided for each bag. A bag label is often only determined by a handful of key instances within a bag, making it difficult to interpret what information a classifier is using to make decisions. In this work, we establish the key requirements for interpreting MIL models. We then go on to develop several model-agnostic approaches that meet these requirements. Our methods are compared against existing inherently interpretable MIL models on several datasets, and achieve an increase in interpretability accuracy of up to 30%. We also examine the ability of the methods to identify interactions between instances and scale to larger datasets, improving their applicability to real-world problems.
2021
Ramchurn, Sarvapali; Middleton, Stuart; McAuley, Derek; Webb, Helena; Hyde, Richard; Lisinska, Justyna
A Response to Draft Online Safety Bill: A call for evidence from the Joint Committee Technical Report
no. 10.18742/pub01-060, 2021.
@techreport{soton451428,
title = {A Response to Draft Online Safety Bill: A call for evidence from the Joint Committee},
author = {Sarvapali Ramchurn and Stuart Middleton and Derek McAuley and Helena Webb and Richard Hyde and Justyna Lisinska},
url = {https://eprints.soton.ac.uk/451428/},
year = {2021},
date = {2021-09-01},
number = {10.18742/pub01-060},
abstract = {This report is the Trustworthy Autonomous Hub (TAS-hub) response to the call for evidence from the Joint Committee on the Draft Online Safety Bill. The Joint Committee was established to consider the Government's draft Bill to establish a new regulatory framework to tackle harmful content online.},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
This report is the Trustworthy Autonomous Hub (TAS-hub) response to the call for evidence from the Joint Committee on the Draft Online Safety Bill. The Joint Committee was established to consider the Government's draft Bill to establish a new regulatory framework to tackle harmful content online.
Ramchurn, Sarvapali; Stein, Sebastian; Jennings, Nicholas R
Trustworthy human-AI partnerships Journal Article
In: iScience, vol. 24, no. 8, 2021.
@article{soton450597,
title = {Trustworthy human-AI partnerships},
author = {Sarvapali Ramchurn and Sebastian Stein and Nicholas R Jennings},
url = {https://eprints.soton.ac.uk/450597/},
year = {2021},
date = {2021-08-01},
journal = {iScience},
volume = {24},
number = {8},
abstract = {In this paper, we foreground some of the key research challenges that arise in the design of trustworthy human-AI partnerships. In particular, we focus on the challenges in designing human-AI partnerships that need to be addressed to help humans and organisations trust their machine counterparts individually or as a collective (e.g., as robot teams or groups of software agents). We also aim to identify the risks associated with human-AI partnerships and therefore determine the associated measures to mitigate these risks. By so doing, we will trigger new avenues of research that will address the key barriers to the adoption of AI-based systems more widely in our daily lives and in industry.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper, we foreground some of the key research challenges that arise in the design of trustworthy human-AI partnerships. In particular, we focus on the challenges in designing human-AI partnerships that need to be addressed to help humans and organisations trust their machine counterparts individually or as a collective (e.g., as robot teams or groups of software agents). We also aim to identify the risks associated with human-AI partnerships and therefore determine the associated measures to mitigate these risks. By so doing, we will trigger new avenues of research that will address the key barriers to the adoption of AI-based systems more widely in our daily lives and in industry.
Ramchurn, Sarvapali; Mousavi, Mohammad Reza; Toliyat, Seyed Mohammad Hossein; Kleinman, Mark; Lisinska, Justyna; Sempreboni, Diego; Stein, Sebastian; Gerding, Enrico; Gomer, Richard; DÁmore, Francesco
The future of connected and automated mobility in the UK: call for evidence Technical Report
no. 10.5258/SOTON/P0097, 2021, (The UKRI TAS Hub assembles a team from the Universities of Southampton, Nottingham and King?s College London. The Hub sits at the centre of the pounds33M Trustworthy Autonomous Systems Programme, funded by the UKRI Strategic Priorities Fund. The role of the TAS Hub is to coordinate and work with six research nodes to establish a collaborative platform for the UK to enable the development of socially beneficial autonomous systems that are both trustworthy in principle and trusted in practice by individuals, society and government. Read more about the TAS Hub at https://www.tas.ac.uk/aboutus/overview/).
@techreport{soton450228,
title = {The future of connected and automated mobility in the UK: call for evidence},
author = {Sarvapali Ramchurn and Mohammad Reza Mousavi and Seyed Mohammad Hossein Toliyat and Mark Kleinman and Justyna Lisinska and Diego Sempreboni and Sebastian Stein and Enrico Gerding and Richard Gomer and Francesco DÁmore},
editor = {Wassim Dbouk},
url = {https://eprints.soton.ac.uk/450228/},
year = {2021},
date = {2021-07-01},
number = {10.5258/SOTON/P0097},
publisher = {University of Southampton},
abstract = {This report is a response to the call for evidence from the Department for Business, Energy & Industrial Strategy and the Centre for Connected and Autonomous Vehicles on the future of connected and automated mobility in the UK.ensuremath<br/ensuremath>Executive Summary:Despite relative weaknesses in global collaboration and co-creation platforms, smart road and communication infrastructure, urban planning, and public awareness, the United Kingdom (UK) has a substantial strength in the area of Connected and Automated Mobility (CAM) by investing in research and innovation platforms for developing the underlying technologies, creating impact, and co-creation leading to innovative solutions. Many UK legal and policymaking initiatives in this domain are world leading. To sustain the UK?s leading position, we make the following recommendations:? The development of financial and policy-related incentive schemes for research and innovation in the foundations and applications of autonomous systems as well as schemes for proof of concepts, and commercialisation.? Supporting policy and standardisation initiatives as well as engagement and community-building activities to increase public awareness and trust.? Giving greater attention to integrating CAM/Connected Autonomous Shared Electric vehicles (CASE) policy with related government priorities for mobility, including supporting active transport and public transport, and improving air quality.? Further investment in updating liability and risk models and coming up with innovative liability schemes covering the Autonomous Vehicles (AVs) ecosystem.? Investing in training and retraining of the work force in the automotive, mobility, and transport sectors, particularly with skills concerningArtificial Intelligence (AI), software and computer systems, in order to ensure employability and an adequate response to the drastically changing industrial landscape},
note = {The UKRI TAS Hub assembles a team from the Universities of Southampton, Nottingham and King?s College London. The Hub sits at the centre of the pounds33M Trustworthy Autonomous Systems Programme, funded by the UKRI Strategic Priorities Fund.
The role of the TAS Hub is to coordinate and work with six research nodes to establish a collaborative platform for the UK to enable the development of socially beneficial autonomous systems that are both trustworthy in principle and trusted in practice by individuals, society and government. Read more about the TAS Hub at https://www.tas.ac.uk/aboutus/overview/},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
This report is a response to the call for evidence from the Department for Business, Energy & Industrial Strategy and the Centre for Connected and Autonomous Vehicles on the future of connected and automated mobility in the UK.ensuremath<br/ensuremath>Executive Summary:Despite relative weaknesses in global collaboration and co-creation platforms, smart road and communication infrastructure, urban planning, and public awareness, the United Kingdom (UK) has a substantial strength in the area of Connected and Automated Mobility (CAM) by investing in research and innovation platforms for developing the underlying technologies, creating impact, and co-creation leading to innovative solutions. Many UK legal and policymaking initiatives in this domain are world leading. To sustain the UK?s leading position, we make the following recommendations:? The development of financial and policy-related incentive schemes for research and innovation in the foundations and applications of autonomous systems as well as schemes for proof of concepts, and commercialisation.? Supporting policy and standardisation initiatives as well as engagement and community-building activities to increase public awareness and trust.? Giving greater attention to integrating CAM/Connected Autonomous Shared Electric vehicles (CASE) policy with related government priorities for mobility, including supporting active transport and public transport, and improving air quality.? Further investment in updating liability and risk models and coming up with innovative liability schemes covering the Autonomous Vehicles (AVs) ecosystem.? Investing in training and retraining of the work force in the automotive, mobility, and transport sectors, particularly with skills concerningArtificial Intelligence (AI), software and computer systems, in order to ensure employability and an adequate response to the drastically changing industrial landscape
Worrawichaipat, Phuriwat; Gerding, Enrico; Kaparias, Ioannis; Ramchurn, Sarvapali
Resilient intersection management with multi-vehicle collision avoidance Journal Article
In: Frontiers in Sustainable Cities, vol. 3, 2021.
@article{soton449675,
title = {Resilient intersection management with multi-vehicle collision avoidance},
author = {Phuriwat Worrawichaipat and Enrico Gerding and Ioannis Kaparias and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449675/},
year = {2021},
date = {2021-06-01},
journal = {Frontiers in Sustainable Cities},
volume = {3},
abstract = {In this paper, we propose a novel decentralised agent-based mechanism for road intersection management for connected autonomous vehicles. In our work we focus on road obstructions causing major traffic delays. In doing so, we propose the first decentralised mechanism able to maximise the overall vehicle throughput at intersections in the presence of obstructions. The distributed algorithm transfers most of the computational cost from the intersection manager to the driving agents, thereby improving scalability. Our realistic empirical experiments using SUMO show that, when an obstacle is located at the entrance or in the middle of the intersection, existing state of the art algorithms and traffic lights show a reduced throughput of 65?90% from the optimal point without obstructions while our mechanism can maintain the throughput upensuremath<br/ensuremath>Q7 to 94?99%.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper, we propose a novel decentralised agent-based mechanism for road intersection management for connected autonomous vehicles. In our work we focus on road obstructions causing major traffic delays. In doing so, we propose the first decentralised mechanism able to maximise the overall vehicle throughput at intersections in the presence of obstructions. The distributed algorithm transfers most of the computational cost from the intersection manager to the driving agents, thereby improving scalability. Our realistic empirical experiments using SUMO show that, when an obstacle is located at the entrance or in the middle of the intersection, existing state of the art algorithms and traffic lights show a reduced throughput of 65?90% from the optimal point without obstructions while our mechanism can maintain the throughput upensuremath<br/ensuremath>Q7 to 94?99%.
Capezzuto, Luca; Tarapore, Danesh; Ramchurn, Sarvapali
Large-scale, dynamic and distributed coalition formation with spatial and†temporal constraints Journal Article
In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), pp. 108–125, 2021.
@article{soton452050,
title = {Large-scale, dynamic and distributed coalition formation with spatial and†temporal constraints},
author = {Luca Capezzuto and Danesh Tarapore and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/452050/},
year = {2021},
date = {2021-05-01},
journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)},
pages = {108–125},
abstract = {The†Coalition Formation with Spatial and Temporal constraints Problem†(CFSTP) is a multi-agent task allocation problem in which few agents have to perform many tasks, each with its deadline and workload. To maximize the number of completed tasks, the agents need to cooperate by forming, disbanding and reforming coalitions. The original mathematical programming formulation of the CFSTP is difficult to implement, since it is lengthy and based on the problematic Big-M method. In this paper, we propose a compact and easy-to-implement formulation. Moreover, we design D-CTS, a distributed version of the state-of-the-art CFSTP algorithm. Using public London Fire Brigade records, we create a dataset with 347588 tasks and a test framework that simulates the mobilization of firefighters in dynamic environments. In problems with up†to 150 agents and 3000 tasks, compared to DSA-SDP, a state-of-the-art distributed algorithm, D-CTS completes†3.79%$pm$[42.22%,1.96%]†more tasks, and is one order of magnitude more efficient in terms of communication overhead and time complexity. D-CTS sets the first large-scale, dynamic and distributed CFSTP benchmark.ensuremath<br/ensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The†Coalition Formation with Spatial and Temporal constraints Problem†(CFSTP) is a multi-agent task allocation problem in which few agents have to perform many tasks, each with its deadline and workload. To maximize the number of completed tasks, the agents need to cooperate by forming, disbanding and reforming coalitions. The original mathematical programming formulation of the CFSTP is difficult to implement, since it is lengthy and based on the problematic Big-M method. In this paper, we propose a compact and easy-to-implement formulation. Moreover, we design D-CTS, a distributed version of the state-of-the-art CFSTP algorithm. Using public London Fire Brigade records, we create a dataset with 347588 tasks and a test framework that simulates the mobilization of firefighters in dynamic environments. In problems with up†to 150 agents and 3000 tasks, compared to DSA-SDP, a state-of-the-art distributed algorithm, D-CTS completes†3.79%$pm$[42.22%,1.96%]†more tasks, and is one order of magnitude more efficient in terms of communication overhead and time complexity. D-CTS sets the first large-scale, dynamic and distributed CFSTP benchmark.ensuremath<br/ensuremath>
Capezzuto, Luca; Tarapore, Danesh; Ramchurn, Sarvapali D
Anytime and efficient multi-agent coordination for disaster response Journal Article
In: SN Computer Science, vol. 2, no. 3, 2021.
@article{soton467373,
title = {Anytime and efficient multi-agent coordination for disaster response},
author = {Luca Capezzuto and Danesh Tarapore and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/467373/},
year = {2021},
date = {2021-03-01},
journal = {SN Computer Science},
volume = {2},
number = {3},
abstract = {The Coalition Formation with Spatial and Temporal constraints Problem (CFSTP) is a multi-agent task allocation problem where the tasks are spatially distributed, with deadlines and workloads, and the number of agents is typically much smaller than the number of tasks. To maximise the number of completed tasks, the agents may have to schedule coalitions. The state-of-the-art CFSTP solver, the Coalition Formation with Look-Ahead (CFLA) algorithm, has two main limitations. First, its time complexity is exponential with the number of agents. Second, as we show, its look-ahead technique is not effective in real-world scenarios, such as open multi-agent systems, where new tasks can appear at any time. In this work, we study its design and define a variant, called Coalition Formation with Improved Look-Ahead (CFLA2), which achieves better performance. Since we cannot eliminate the limitations of CFLA in CFLA2, we also develop a novel algorithm to solve the CFSTP, the first to be simultaneously anytime, efficient and with convergence guarantee, called Cluster-based Task Scheduling (CTS). In tests where the look-ahead technique is highly effective, CTS completes up to 30% (resp. 10%) more tasks than CFLA (resp. CFLA2) while being up to 4 orders of magnitude faster. We also propose S-CTS, a simplified but parallel variant of CTS with even lower time complexity. Using scenarios generated by the RoboCup Rescue Simulation, we show that S-CTS is at most 10% less performing than high-performance algorithms such as Binary Max-Sum and DSA, but up to 2 orders of magnitude faster. Our results affirm CTS as the new state-of-the-art algorithm to solve the CFSTP.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The Coalition Formation with Spatial and Temporal constraints Problem (CFSTP) is a multi-agent task allocation problem where the tasks are spatially distributed, with deadlines and workloads, and the number of agents is typically much smaller than the number of tasks. To maximise the number of completed tasks, the agents may have to schedule coalitions. The state-of-the-art CFSTP solver, the Coalition Formation with Look-Ahead (CFLA) algorithm, has two main limitations. First, its time complexity is exponential with the number of agents. Second, as we show, its look-ahead technique is not effective in real-world scenarios, such as open multi-agent systems, where new tasks can appear at any time. In this work, we study its design and define a variant, called Coalition Formation with Improved Look-Ahead (CFLA2), which achieves better performance. Since we cannot eliminate the limitations of CFLA in CFLA2, we also develop a novel algorithm to solve the CFSTP, the first to be simultaneously anytime, efficient and with convergence guarantee, called Cluster-based Task Scheduling (CTS). In tests where the look-ahead technique is highly effective, CTS completes up to 30% (resp. 10%) more tasks than CFLA (resp. CFLA2) while being up to 4 orders of magnitude faster. We also propose S-CTS, a simplified but parallel variant of CTS with even lower time complexity. Using scenarios generated by the RoboCup Rescue Simulation, we show that S-CTS is at most 10% less performing than high-performance algorithms such as Binary Max-Sum and DSA, but up to 2 orders of magnitude faster. Our results affirm CTS as the new state-of-the-art algorithm to solve the CFSTP.
Ortega, Andre P; Ramchurn, Sarvapali; Tran-Thanh, Long; Merrett, Geoff
Partner selection in self-organised wireless sensor networks for opportunistic energy negotiation: A multi-armed bandit based approach Journal Article
In: Ad Hoc Networks, vol. 112, 2021.
@article{soton445733,
title = {Partner selection in self-organised wireless sensor networks for opportunistic energy negotiation: A multi-armed bandit based approach},
author = {Andre P Ortega and Sarvapali Ramchurn and Long Tran-Thanh and Geoff Merrett},
url = {https://eprints.soton.ac.uk/445733/},
year = {2021},
date = {2021-03-01},
journal = {Ad Hoc Networks},
volume = {112},
abstract = {The proliferation of ?Things? over a network creates the Internet of Things (IoT), where sensors integrate to collect data from the environment over long periods of time. The growth of IoT applications will inevitably involve co-locating multiple wireless sensor networks, each serving different applications with, possibly, different needs and constraints. Since energy is scarce in sensor nodes equipped with non-rechargeable batteries, energy harvesting technologies have been the focus of research in recent years. However, new problems arise as a result of their wide spatio-temporal variation. Such a shortcoming can be avoided if co-located networks cooperate with each other and share their available energy. Due to their unique characteristics and different owners, recently, we proposed a negotiation approach to deal with conflict of preferences. Unfortunately, negotiation can be impractical with a large number of participants, especially in an open environment. Given this, we introduce a new partner selection technique based on multi-armed bandits (MAB), that enables each node to learn the strategy that optimises its energy resources in the long term. Our results show that the proposed solution allows networks to repeatedly learn the current best energy partner in a dynamic environment. The performance of such a technique is evaluated through simulation and shows that a network can achieve an efficiency of 72% against the optimal strategy in the most challenging scenario studied in this work.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The proliferation of ?Things? over a network creates the Internet of Things (IoT), where sensors integrate to collect data from the environment over long periods of time. The growth of IoT applications will inevitably involve co-locating multiple wireless sensor networks, each serving different applications with, possibly, different needs and constraints. Since energy is scarce in sensor nodes equipped with non-rechargeable batteries, energy harvesting technologies have been the focus of research in recent years. However, new problems arise as a result of their wide spatio-temporal variation. Such a shortcoming can be avoided if co-located networks cooperate with each other and share their available energy. Due to their unique characteristics and different owners, recently, we proposed a negotiation approach to deal with conflict of preferences. Unfortunately, negotiation can be impractical with a large number of participants, especially in an open environment. Given this, we introduce a new partner selection technique based on multi-armed bandits (MAB), that enables each node to learn the strategy that optimises its energy resources in the long term. Our results show that the proposed solution allows networks to repeatedly learn the current best energy partner in a dynamic environment. The performance of such a technique is evaluated through simulation and shows that a network can achieve an efficiency of 72% against the optimal strategy in the most challenging scenario studied in this work.
Yazdanpanah, Vahid; Gerding, Enrico H; Stein, Sebastian; Dastani, Mehdi; Jonker, Catholijn M; Norman, Timothy J; Ramchurn, Sarvapali D
Responsibility ascription in trustworthy autonomous systems Proceedings Article
In: Embedding AI in Society (18/02/21 - 19/02/21), 2021.
@inproceedings{soton446459,
title = {Responsibility ascription in trustworthy autonomous systems},
author = {Vahid Yazdanpanah and Enrico H Gerding and Sebastian Stein and Mehdi Dastani and Catholijn M Jonker and Timothy J Norman and Sarvapali D Ramchurn},
url = {https://eprints.soton.ac.uk/446459/},
year = {2021},
date = {2021-02-01},
booktitle = {Embedding AI in Society (18/02/21 - 19/02/21)},
abstract = {To develop and effectively deploy Trustworthy Autonomous Systems (TAS), we face various social, technological, legal, and ethical challenges in which different notions of responsibility can play a key role. In this work, we elaborate on these challenges, discuss research gaps, and show how the multidimensional notion of responsibility can play a key role to bridge them. We argue that TAS requires operational tools to represent and reason about the responsibilities of humans as well as AI agents.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
To develop and effectively deploy Trustworthy Autonomous Systems (TAS), we face various social, technological, legal, and ethical challenges in which different notions of responsibility can play a key role. In this work, we elaborate on these challenges, discuss research gaps, and show how the multidimensional notion of responsibility can play a key role to bridge them. We argue that TAS requires operational tools to represent and reason about the responsibilities of humans as well as AI agents.
Beal, Ryan James; Middleton, Stuart; Norman, Timothy; Ramchurn, Sarvapali
Combining machine learning and human experts to predict match outcomes in football: A baseline model Proceedings Article
In: The Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (02/02/21 - 09/02/21), 2021.
@inproceedings{soton445607,
title = {Combining machine learning and human experts to predict match outcomes in football: A baseline model},
author = {Ryan James Beal and Stuart Middleton and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/445607/},
year = {2021},
date = {2021-02-01},
booktitle = {The Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (02/02/21 - 09/02/21)},
abstract = {In this paper, we present a new application-focused benchmark dataset and results from a set of baseline Natural Language Processing and Machine Learning models for prediction of match outcomes for games of football (soccer). By doing so we give a baseline for the prediction accuracy that can be achieved exploiting both statistical match data and contextual articles from human sports journalists. Our dataset is focuses on a representative time-period over 6 seasons of the English Premier League, and includes newspaper match previews from The Guardian. The models presented in this paper achieve an accuracy of 63.18% showing a 6.9% boost on the traditional statistical methods.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we present a new application-focused benchmark dataset and results from a set of baseline Natural Language Processing and Machine Learning models for prediction of match outcomes for games of football (soccer). By doing so we give a baseline for the prediction accuracy that can be achieved exploiting both statistical match data and contextual articles from human sports journalists. Our dataset is focuses on a representative time-period over 6 seasons of the English Premier League, and includes newspaper match previews from The Guardian. The models presented in this paper achieve an accuracy of 63.18% showing a 6.9% boost on the traditional statistical methods.
Lhopital, Sacha; Aknine, Samir; Ramchurn, Sarvapali; Thavonekham, Vincent; Vu, Huan
Decentralised control of intelligent devices: a healthcare facility study Proceedings Article
In: Bassiliades, Nick; Chalkiadakis, Georgios; Jonge, Dave (Ed.): Multi-Agent Systems and Agreement Technologies - 17th European Conference, EUMAS 2020, and 7th International Conference, AT 2020, Revised Selected Papers, pp. 20–36, Springer, 2021.
@inproceedings{soton447983,
title = {Decentralised control of intelligent devices: a healthcare facility study},
author = {Sacha Lhopital and Samir Aknine and Sarvapali Ramchurn and Vincent Thavonekham and Huan Vu},
editor = {Nick Bassiliades and Georgios Chalkiadakis and Dave Jonge},
url = {https://eprints.soton.ac.uk/447983/},
year = {2021},
date = {2021-01-01},
booktitle = {Multi-Agent Systems and Agreement Technologies - 17th European Conference, EUMAS 2020, and 7th International Conference, AT 2020, Revised Selected Papers},
volume = {12520 LNAI},
pages = {20–36},
publisher = {Springer},
abstract = {ensuremath<pensuremath>We present a novel approach to the management of notifications from devices in a healthcare setting. We employ a distributed constraint optimisation (DCOP) approach to the delivery of notification for healthcare assistants that aims to preserve the privacy of patients while reducing the intrusiveness of such notifications. Our approach reduces the workload of the assistants and improves patient safety by automating task allocation while ensuring high priority needs are addressed in a timely manner. We propose and evaluate several DCOP models both in simulation and in real-world deployments. Our models are shown to be efficient both in terms of computation and communication costs.ensuremath</pensuremath>},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
ensuremath<pensuremath>We present a novel approach to the management of notifications from devices in a healthcare setting. We employ a distributed constraint optimisation (DCOP) approach to the delivery of notification for healthcare assistants that aims to preserve the privacy of patients while reducing the intrusiveness of such notifications. Our approach reduces the workload of the assistants and improves patient safety by automating task allocation while ensuring high priority needs are addressed in a timely manner. We propose and evaluate several DCOP models both in simulation and in real-world deployments. Our models are shown to be efficient both in terms of computation and communication costs.ensuremath</pensuremath>
Ryan, James Beal; Chalkiadakis, Georgios; Norman, Timothy; Ramchurn, Sarvapali
Optimising long-term outcomes using real-world fluent objectives: an application to football Proceedings Article
In: 20th International Conference on Autonomous Agents and Multiagent Systems (03/05/21 - 07/05/21), pp. 196–204, 2021.
@inproceedings{soton449655,
title = {Optimising long-term outcomes using real-world fluent objectives: an application to football},
author = {James Beal Ryan and Georgios Chalkiadakis and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449655/},
year = {2021},
date = {2021-01-01},
booktitle = {20th International Conference on Autonomous Agents and Multiagent Systems (03/05/21 - 07/05/21)},
pages = {196–204},
abstract = {In this paper, we present a novel approach for optimising long-term tactical and strategic decision-making in football (soccer) by encapsulating events in a league environment across a given time frame. ensuremath<br/ensuremath>We model the teams' objectives for a season and track how these evolve as games unfold to give a fluent objective that can aid in decision-making games. We develop Markov chain Monte Carlo and deep learning-based algorithms that make use of the fluent objectives in order to learn from prior games and other games in the environment and increase the teams? long-term performance. Simulations of our approach using real-world datasets from 760 matches shows that by using optimised tactics with our fluent objective and prior games, we can on average increase teams mean expected finishing distribution in the league by up to 35.6%.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we present a novel approach for optimising long-term tactical and strategic decision-making in football (soccer) by encapsulating events in a league environment across a given time frame. ensuremath<br/ensuremath>We model the teams' objectives for a season and track how these evolve as games unfold to give a fluent objective that can aid in decision-making games. We develop Markov chain Monte Carlo and deep learning-based algorithms that make use of the fluent objectives in order to learn from prior games and other games in the environment and increase the teams? long-term performance. Simulations of our approach using real-world datasets from 760 matches shows that by using optimised tactics with our fluent objective and prior games, we can on average increase teams mean expected finishing distribution in the league by up to 35.6%.
Merhej, Charbel; Ryan, James Beal; Matthews, Tim; Ramchurn, Sarvapali
What happened next? Using deep learning to value defensive actions in football event-data Proceedings Article
In: KDD 2021 (14/08/21 - 18/08/21), pp. 3394–3403, 2021.
@inproceedings{soton449656,
title = {What happened next? Using deep learning to value defensive actions in football event-data},
author = {Charbel Merhej and James Beal Ryan and Tim Matthews and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/449656/},
year = {2021},
date = {2021-01-01},
booktitle = {KDD 2021 (14/08/21 - 18/08/21)},
pages = {3394–3403},
abstract = {Objectively quantifying the value of player actions in football (soccer) is a challenging problem. To date, studies in football analytics have mainly focused on the attacking side of the game, while there has been less work on event-driven metrics for valuing defensive actions (e.g., tackles and interceptions). Therefore in this paper, we use deep learning techniques to define a novel metric that values such defensive actions by studying the threat of passages of play that preceded them. By doing so, we are able to value defensive actions based on what they prevented from happening in the game. Our Defensive Action Expected Threat (DAxT) model has been validated using real-world event-data from the 2017/2018 and 2018/2019 English Premier League seasons, and we combine our model outputs with additional features to derive an overall rating of defensive ability for players. Overall, we find that our model is able to predict the impact of defensive actions allowing us to better value defenders using event-data.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Objectively quantifying the value of player actions in football (soccer) is a challenging problem. To date, studies in football analytics have mainly focused on the attacking side of the game, while there has been less work on event-driven metrics for valuing defensive actions (e.g., tackles and interceptions). Therefore in this paper, we use deep learning techniques to define a novel metric that values such defensive actions by studying the threat of passages of play that preceded them. By doing so, we are able to value defensive actions based on what they prevented from happening in the game. Our Defensive Action Expected Threat (DAxT) model has been validated using real-world event-data from the 2017/2018 and 2018/2019 English Premier League seasons, and we combine our model outputs with additional features to derive an overall rating of defensive ability for players. Overall, we find that our model is able to predict the impact of defensive actions allowing us to better value defenders using event-data.
2020
Beal, Ryan James; Norman, Timothy; Ramchurn, Sarvapali
Optimising daily fantasy sports teams with artificial intelligence Journal Article
In: International Journal of Computer Science in Sport, vol. 19, no. 2, 2020.
@article{soton445995,
title = {Optimising daily fantasy sports teams with artificial intelligence},
author = {Ryan James Beal and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/445995/},
year = {2020},
date = {2020-12-01},
journal = {International Journal of Computer Science in Sport},
volume = {19},
number = {2},
abstract = {This paper outlines a novel approach to optimising teams for Daily Fantasy Sports (DFS) contests. To this end, we propose a number of new models and algorithms to solve the team formation problems posed by DFS. Specifically, we focus on the National Football League (NFL) and predict the performance of real-world players to form the optimal fantasy team using mixed-integer programming. We test our solutions using real-world data-sets from across four seasons (2014-2017). We highlight the advantage that can be gained from using our machine-based methods and show that our solutions outperform existing benchmarks, turning a profit in up to 81.3% of DFS game-weeks over a season},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper outlines a novel approach to optimising teams for Daily Fantasy Sports (DFS) contests. To this end, we propose a number of new models and algorithms to solve the team formation problems posed by DFS. Specifically, we focus on the National Football League (NFL) and predict the performance of real-world players to form the optimal fantasy team using mixed-integer programming. We test our solutions using real-world data-sets from across four seasons (2014-2017). We highlight the advantage that can be gained from using our machine-based methods and show that our solutions outperform existing benchmarks, turning a profit in up to 81.3% of DFS game-weeks over a season
Beal, Ryan James; Norman, Timothy; Ramchurn, Sarvapali
A critical comparison of machine learning classifiers to predict match outcomes in the NFL Journal Article
In: International Journal of Computer Science in Sport, vol. 19, no. 2, 2020.
@article{soton446078,
title = {A critical comparison of machine learning classifiers to predict match outcomes in the NFL},
author = {Ryan James Beal and Timothy Norman and Sarvapali Ramchurn},
url = {https://eprints.soton.ac.uk/446078/},
year = {2020},
date = {2020-12-01},
journal = {International Journal of Computer Science in Sport},
volume = {19},
number = {2},
abstract = {In this paper, we critically evaluate the performance of nine machine learning classification techniques when applied to the match outcome prediction problem presented by American Football. Specifically, we implement and test nine techniques using real-world datasets of 1280 games over 5 seasons from the National Football League (NFL). We test the nine different classifier techniques using a total of 42 features for each team and we find that the best performing algorithms are able to improve one previous published works. The algoriothms achieve an accuracy of between 44.64% for a Guassian Process classifier to 67.53% with a Na"ive Bayes classifer. We also test each classifier on a year by year basis and compare our results to those of the bookmakers and other leading academic papers.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper, we critically evaluate the performance of nine machine learning classification techniques when applied to the match outcome prediction problem presented by American Football. Specifically, we implement and test nine techniques using real-world datasets of 1280 games over 5 seasons from the National Football League (NFL). We test the nine different classifier techniques using a total of 42 features for each team and we find that the best performing algorithms are able to improve one previous published works. The algoriothms achieve an accuracy of between 44.64% for a Guassian Process classifier to 67.53% with a Na"ive Bayes classifer. We also test each classifier on a year by year basis and compare our results to those of the bookmakers and other leading academic papers.
Rigas, Emmanouil S; Gerding, Enrico; Stein, Sebastian; Ramchurn, Sarvapali D; Bassiliades, Nick
Mechanism design for efficient allocation of electric vehicles to charging stations Proceedings Article
In: SETN 2020: 11th Hellenic Conference on Artificial Intelligence, pp. 10–15, 2020.
@inproceedings{soton446412,
title = {Mechanism design for efficient allocation of electric vehicles to charging stations},
author = {Emmanouil S Rigas and Enrico Gerding and Sebastian Stein and Sarvapali D Ramchurn and Nick Bassiliades},
url = {https://eprints.soton.ac.uk/446412/},
year = {2020},
date = {2020-09-01},
booktitle = {SETN 2020: 11th Hellenic Conference on Artificial Intelligence},
pages = {10–15},
abstract = {The electrification of transport can significantly reduce CO2 emissions and their negative impact on the environment. In this paper, we study the problem of allocating Electric Vehicles (EVs) to charging stations and scheduling their charging. We develop an offline solution that treats EV users as self-interested agents that aim to maximise their profit and minimise the impact on their schedule. We formulate the problem of the optimal EV to charging station allocation as a Mixed Integer Programming (MIP) one and we propose two pricing mechanisms: A fixed-price one, and another that is based on the well known Vickrey-Clark-Groves (VCG) mechanism. We observe that the VCG mechanism services on average 1.5% more EVs than the fixed-price one. In addition, when the stations get congested, VCG leads to higher prices for the EVs and higher profit for the stations, but lower utility for the EVs. However, the VCG mechanism guarantees truthful reporting of the EVs? preferences.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The electrification of transport can significantly reduce CO2 emissions and their negative impact on the environment. In this paper, we study the problem of allocating Electric Vehicles (EVs) to charging stations and scheduling their charging. We develop an offline solution that treats EV users as self-interested agents that aim to maximise their profit and minimise the impact on their schedule. We formulate the problem of the optimal EV to charging station allocation as a Mixed Integer Programming (MIP) one and we propose two pricing mechanisms: A fixed-price one, and another that is based on the well known Vickrey-Clark-Groves (VCG) mechanism. We observe that the VCG mechanism services on average 1.5% more EVs than the fixed-price one. In addition, when the stations get congested, VCG leads to higher prices for the EVs and higher profit for the stations, but lower utility for the EVs. However, the VCG mechanism guarantees truthful reporting of the EVs? preferences.
Oluwasuji, Olabambo Ifeoluwa; Malik, Obaid; Zhang, Jie; Ramchurn, Sarvapali Dyanand
Solving the fair electric load shedding problem in developing countries Journal Article
In: Autonomous Agents and Multi-Agent Systems, vol. 34, no. 1, pp. 12, 2020.
@article{oluwasuji2020solving,
title = {Solving the fair electric load shedding problem in developing countries},
author = {Olabambo Ifeoluwa Oluwasuji and Obaid Malik and Jie Zhang and Sarvapali Dyanand Ramchurn},
year = {2020},
date = {2020-01-01},
journal = {Autonomous Agents and Multi-Agent Systems},
volume = {34},
number = {1},
pages = {12},
publisher = {Springer},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Oluwasuji, Olabambo Ifeoluwa; Malik, Obaid; Zhang, Jie; Ramchurn, Sarvapali Dyanand
Solving the fair electric load shedding problem in developing countries Journal Article
In: Auton. Agents Multi Agent Syst., vol. 34, no. 1, pp. 12, 2020.
@article{DBLP:journals/aamas/OluwasujiMZR20,
title = {Solving the fair electric load shedding problem in developing countries},
author = {Olabambo Ifeoluwa Oluwasuji and Obaid Malik and Jie Zhang and Sarvapali Dyanand Ramchurn},
url = {https://doi.org/10.1007/s10458-019-09428-8},
doi = {10.1007/s10458-019-09428-8},
year = {2020},
date = {2020-01-01},
journal = {Auton. Agents Multi Agent Syst.},
volume = {34},
number = {1},
pages = {12},
keywords = {},
pubstate = {published},
tppubtype = {article}
}