Newsletter Issue - 02-03-2026

CDT Weekly Newsletter


Welcome to this week’s issue of the newsletter.



Message from the Director 


We had a very enjoyable graduation day on Wednesday, with a record 10 IAI students graduating! Many congratulations to Drs Whettam (CDT2019), Brookes, Deane, Flanagan, Hanslope, Sahota, Turco (CDT2020), Davies, Degen and Nzoyem (CDT2021). 



CDT AI Symposium – Foundation and Applications of Trustworthy AI 


Date: 27 March, 9:30 – 16:00 

Location: MVB 1.11A and Lower Atrium 

Registration deadline: Monday 16 March 

📢 Save the date! This symposium aims to advance understanding of Trustworthy AI by connecting foundational methods and emerging applications, while fostering discussion on how rapid advances in AI are reshaping scientific practice itself.  

More information including the programme and registration link can be found on the AI Symposium website here.  



AI Lunch and Learn Seminars


Nan Lu: When Data Lies: Building Trustworthy AI from Imperfect Information 

Date: 4 March, pizza from 13:30 – 14:00 followed by the talk from 14:00 – 15:00 

Location: Room 1.07, Queen’s Building 

Modern AI thrives on large models and massive datasets. Yet in many real-world domains, from healthcare records to social systems, labels are often missing, biased, or unreliable, and data evolves in messy, unpredictable ways. This raises a fundamental question: how can we build AI we can trust when the data itself cannot be trusted?  

Abstract: In this talk, I’ll take you behind the scenes of our research tackling exactly this challenge. You’ll see how we can: 

  • Train models without explicit labels; 

  • Adapt to shifting data distributions; 

  • Enable multiple organisations to collaborate on training a federated model using only unlabelled, decentralised data; 

  • Improve reinforcement learning agents even when reward signals are unreliable. 

 

We’ll wrap up by exploring new ways to evaluate AI, going beyond simple accuracy to “health checkups” for models that ensure they remain reliable and robust in the wild. Whether you’re curious about theory, algorithms, or real-world impact, this talk aims to inspire fresh ideas for thinking differently about AI research. 

Bio: Dr. Nan Lu is a Lecturer in Artificial Intelligence in the School of Computer Science at the University of Bristol. She was a postdoctoral fellow in the Foundations of Machine Learning Systems Group at the University of Tübingen, Germany, working with Prof. Robert Williamson, and obtained her Ph.D. in Machine Learning from the University of Tokyo under the supervision of Prof. Masashi Sugiyama. Her research focuses on trustworthy AI, developing principled algorithms that remain reliable under data corruption and across diverse modalities, with applications in core AI areas such as computer vision and reinforcement learning, as well as real-world problems in healthcare and social data. 



Digital Health Seminar Series 


  1. Mengyan Zhang: Learning to Decide Under Uncertainty: From Bandit Theory to Health Applications 

Date: 03 March 14:00 – 15:00 

Location: F40 Lecture Theatre, Biomedical Science Building (University of Bristol only) & Online (TicketTailor link) 

Abstract: How can AI systems make reliable sequential decisions when interacting with dynamic environments under limited budgets? Sequential decision-making frameworks such as multi-armed bandits, reinforcement learning, and active learning provide principled foundations for addressing these challenges in health and science. 

In this talk, I will present sequential decision problems arising in biological sequence modelling and discovery, as well as disease surveillance using graph-based active learning. These examples illustrate how learning under uncertainty can guide data collection, experimentation, and intervention in resource-constrained health contexts. I will conclude by outlining open challenges and future research directions at the intersection of sequential decision-making and digital health. 

 

  1. Carme Carrion Ribas: Measuring What Matters: Why Assessment is the Cornerstone of Digital Health Success 

Date: 17 March 14:00 – 15:00 

Location: In Person at Arts Complex B71 (University of Bristol only) & Online Teams link 

Abstract: The digital transformation of healthcare holds the promise of more efficient and patient-centered systems. Yet, many promising digital health interventions fail to move beyond the “pilot phase” due to a lack of robust evidence. This presentation argues that comprehensive assessment is not merely a final step, but the cornerstone of digital health sustainability. We will examine the evaluation lifecycle, starting with the fundamental assessment of usability and feasibility to ensure technological maturity. However, the shift from a “standalone product” to a “healthcare intervention” requires rigorous validation in real-world clinical settings to guarantee safety, security, and clinical efficacy. Crucially, as we look toward large-scale implementation, the scope of assessment must broaden. We will discuss the necessity of evaluating organizational readiness, ethical implications, economic viability, and—increasingly vital—environmental impact. By integrating these perspectives, we can move toward a holistic framework that considers the needs of all stakeholders: from the individual citizen and healthcare provider to the broader health system and society. 



Forthcoming Calendar of Events



AI Lunch and Learn Seminars