When can we trust an AI decision?
19 October 2026
The next Coda Network webinar is due to take place on Tuesday 27 October at 12 noon and will look at AI decision-making and issues of trust. Full details are given below. All are welcome; if you don’t already have the recurring invitation to the webinars, contact us if you want to take part.When Can We Trust an AI Decision? Reliability, Assurance and Accountability in Health and Finance
AI is moving from advising people to acting on their behalf, in clinics, banks and beyond. But a confident answer is not a justified one. The most dangerous failures rarely come from a lack of intelligence. They come from misleading similarity: new cases that look like the data a model learned from, yet differ in the ways that matter, such as a rare patient or a sudden shift in market conditions.
In this talk Dr Zhongtian Sun (Kent) will show how such failures can be caught at three levels:
- Inside the model: geometry and topology reveal when a network has memorised rather than learned. They also show why even a trust signal must be validated in the setting where it is used.
- At decision time: “self-certifying” systems monitor themselves and respond in proportion: proceed, flag, defer to a human, or abstain.
- Around the system: the Infrastructure–Intent–Deeds framework holds AI agents to account for what they actually did.
Speaker
Dr Zhongtian Sun is a Lecturer in AI at the School of Computing, University of Kent, and a Visiting Academic at the University of Cambridge. His research develops trustworthy and reliable AI for high-stakes decisions, drawing on geometric and topological deep learning, with applications in healthcare and finance. His work has been published at leading venues including NeurIPS, ICML, The Web Conference, ACM RecSys and ICAIF. He was the inaugural AI Policy Fellow at Mila – Quebec AI Institute, and contributes to the European Commission AI Office Expert Forum on Frontier AI and the OECD Working Group on Agentic AI. He is Competition Chair for ICAIF 2026. Before academia, he co-founded the AI fintech start-up AI4Finance London. He holds a PhD from Durham University.