Summer School 2025
This year’s Safe and Trusted AI Summer School is co-organised by the UKRI Centre for Doctoral Training in Safe and Trusted AI (STAI CDT) and the UKRI Centre for Doctoral Training in Machine Intelligence for Nano-Electronic Devices and Systems (MINDS CDT).
Speakers
We have an exciting line up of speakers, from both academia and industry. Confirmed speakers include:
Prof Kate Devlin, Professor of AI & Society, King’s College London
Professor Devlin’s research investigates how – and why – people interact with and react to technologies, both past and future. She is the author of the critically acclaimed book “Turned On: Science, Sex and Robots” (Bloomsbury, 2018), which examines the ethical and social implications of technology and intimacy. She is Chair-Director of the Digital Futures Institute at King’s, and is King’s investigator on the UKRI’s £31 million Responsible AI UK programme, which brings together researchers from across the UK to understand how to shape the development of AI to benefit people, communities and society.
Prof David Parker, Professor of Computer Science, University of Oxford
Professor Parker’s research is in formal verification: rigorous techniques for checking that systems function correctly. In particular, he works on quantitative verification methods for checking properties such as safety, reliability, robustness and performance. Professor Parker leads the development of PRISM, the most widely-used software tool for verification of probabilistic systems.
Dr Chenxi Whitehouse, Research Scientist, Meta
Dr Chenxi Whitehouse is a research scientist at Meta, where she focuses on the evaluation of Large Language Models. She is also a visiting researcher at the University of Cambridge, where she previously worked as a postdoctoral research associate on automated fact-checking. Before joining Meta, Chenxi was an applied research scientist at Amazon AGI. She holds a PhD in knowledge-grounded natural language processing from City, University of London. Chenxi is an active contributor to the NLP community, with publications and reviewing experience at top venues including ACL, EMNLP, and NeurIPS.
Dr Edgar Lopez-Rojas, Founder and CEO of RevAIsor
Dr. Edgar Lopez-Rojas is the CEO and Founder of RevAIsor, a London-based company that aims to empower positive change with trustworthy and ethical AI. He holds a PhD in Computer Science from BTH in Sweden. He has over 20 years of experience in software development, data analytics, and computational methods for Security and Financial Crime Analytics.
Dr. Lopez-Rojas is also the visionary Founder and Chief Visionary Officer of FinCrime Dynamics, a company specialising in Simulation for Financial Crime Analytics using Synthetic Data. He invented the “FinCrime Vaccines” concept, a groundbreaking approach to enhancing Anti-Money Laundering Compliance and Optimisation within the financial industry. He is an Innovate UK awardee, a prominent international speaker, and a distinguished researcher who pioneered methods to synthesise financial transactions and measure the performance and effectiveness of FinCrime analytics. He is passionate about harnessing the potential of cutting-edge technologies and innovative approaches to address the risks and challenges posed by AI biases, ethical concerns, and regulatory complexities.
Dr Stuart E. Middleton, Associate Professor, University of Southampton
Dr Stuart Middleton is an Associate Professor at the University of Southampton. He has more than 60 peer reviewed publications, many inter-disciplinary in nature, focussing on the Natural Language Processing (NLP) areas of information extraction and human-in-the-loop NLP. His research interests are focussed on socio-technical NLP approaches, including Large Language Models (LLMs), few/zero-shot learning, rationale-based learning, adversarial training and argument mining. He has been an PI/CoI on grants valued over £47M total (£2.29M income for University of Southampton), leading inter-disciplinary multi-organisational projects in domains such as law enforcement, defence and security, mental health and environmental science. He is deputy director of the £5.8M MINDS Centre for Doctoral Training, visiting researcher at Northeastern University, Turing Fellow (2021 – 2023) and board member of the Centre for Machine Intelligence. He has served as an invited expert at various events including UK Cabinet Office Ministerial AI Roundtable 2019 on ‘Use of AI in Policing’.
Prof Amanda Prorok, Professor of Collective Intelligence and Robotics, University of Cambridge
Professor Prorok’s lab at the University of Cambridge develops solutions for collective intelligence in multi-agent and multi-robot systems. Their mission is to find new ways of coordinating artificially intelligent agents (e.g., robots, vehicles, machines) to achieve common goals in shared physical and virtual spaces. Together with her lab, Professor Prorok pioneered methods for differentiable communication between learning agents.
Registration Information
The cost of registration is £200, which includes food and beverages for coffee and lunch breaks across the three days.
Register for your place at:
STAI summer school 2025 | Imperial College London Online Store
Registrations for external participants close on Friday 20 June at 5pm.
Schedule
The Summer School will take place over three days, 14-16 July 2025 at Imperial College London.
Please see the schedule below for timings:
| Day 1: Mon 14 July | Day 2: Tues 15 July | Day 3: Wed 16 July | |||
| 10:15 – 10:30 | Introduction to Safe and Trusted AI | 09:30-10:45 | Amanda Prorok (P.1) | 09:30-10:45 | Kate Devlin (P.1) |
| 10:30-11:30 | Chenxi Whitehouse (P.1) | 10:45-11:00 | Break | 10:45-11:00 | Break |
| 11:30-11:45 | Break | 11:00-12:15 | Amanda Prorok (P.2) | 11:00-12:15 | Kate Devlin (P.2) |
| 11:45-13:15 | Chenxi Whitehouse (P.2) | 12:15-13:45 | Lunch Break + Poster Session | 12:15-13:45 | Lunch Break+ Poster Session |
| 13:15-14:15 | Lunch Break + Poster Session | 13:45-15:00 | Stuart Middleton (P.1) | 13:45-15:00 | David Parker (P.1) |
| 14:15-15:30 | Edgar Lopez-Rojas (P.1) | 15:00-15:15 | Break | 15:00-15:15 | Break |
| 15:30-15:45 | Break | 15:15-16:30 | Stuart Middleton (P.2) | 15:15-16:30 | David Parker (P.2) |
| 15:45-17:00 | Edgar Lopez-Rojas (P.2) | 16:30-18:30 | Social Reception | 16:30-17:00 | Closing Remarks |
Talks and abstracts:
Prof Kate Devlin – Without interdisciplinarity there can be no responsibility
Responsible AI is not solely a technical challenge: it is also a social one, a cultural one, and a human one. Designing, developing, and deploying AI in ways that are ethical, inclusive, and accountable requires the combined insight of diverse disciplines. No single field can hold the answers to the complex challenge: we need to work together across boundaries, using a variety of methodologies and a combination of approaches, to anticipate risks, mitigate harms, and ensure AI serves the public good. Interdisciplinary collaboration is not optional – it is foundational to building systems that reflect pluralistic values. It’s also not easy – and this talk will outline the pitfalls and propose good practices.
Dr Chenxi Whitehouse – Incentivizing Thinking in LLM-as-a-Judge and Reward Modelling via Reinforcement Learning
As large language models (LLMs) become increasingly central to AI systems, reliable evaluation of their outputs has emerged as a critical challenge. Recent advances demonstrate that LLMs themselves can serve as effective evaluators—or “LLM-as-a-Judge”—particularly when trained to reason systematically. This talk examines how reinforcement learning (RL) can enhance reasoning depth in evaluator models.
The session will begin with an analysis of current LLM-based evaluation approaches and their limitations such as in addressing judgment biases. The presentation will cover a progression of work including Self-Taught Evaluators and EvalPlanner before introducing the J1 framework. This approach applies unified reinforcement learning to train judgment models using verifiable rewards that incentivize chain-of-thought reasoning.
The J1 framework achieves state-of-the-art results across multiple benchmarks, outperforming larger models like DeepSeek-R1 and o1-mini, even at 8B and 70B parameter scales. The talk will explore key ablation studies—including pairwise versus pointwise training, online versus offline learning, and the impact of reward design, prompt structure, and reasoning length on evaluation quality. The talk will conclude with future directions, including how LLM-as-a-Judge can be better used for Reward Modelling in post-training.
Dr Edgar Lopez-Rojas – Testing Trustworthy AI: Bridging Regulation, Research Frontiers, and Practical Implementation Challenges
The pursuit of trustworthy Artificial Intelligence hinges on rigorous and innovative testing practices. This presentation surveys the key elements involved in achieving reliable AI systems. It will cover the impact of current and forthcoming AI regulations on testing requirements, with attention to the evolving European framework. We will then review practical testing techniques, showcasing the utility of “LLM-as-a-Judge” for sophisticated output evaluation and “Agent-Based Simulation” for understanding AI in complex interactive environments. The presentation aims to map out the terrain of existing work and, more importantly, to illuminate compelling research opportunities for PhD students looking to contribute to this essential and rapidly advancing field.
Dr Stuart E. Middleton – Large Language Models (LLMs) and Human-in-the-loop NLP: Tutorial and Research Insights
This session will provide a tutorial on Large Language Models (LLMs) and Human-in-the-loop NLP. It will cover what LLMs are, LLM developments leading to the powerful LLMs we see today and what Human-in-the-loop NLP methodologies are available to use with them. Insights will then be reported from recent research into LLMs and Human-in-the-loop NLP focussing on critical application domains such as defence, law enforcement and mental health, where getting an answer right in a safe and trustworthy way really matters.