Among the presenters at a major ACM conference today will be an undergraduate from this Department, sharing research on ways to improve bias and accuracy in new AI tools that are being developed to assess patients' mental health.

Tom Brennan is a current Part II student here. At the ACM International Conference On Multimodal Interaction today (6 October 2026), he will discuss 'FAIR_XAI: Improving Multimodal Foundation Model Fairness via Explainability for Wellbeing Assessment'.

This is a paper he co-authored earlier this year with fellow Computer Science undergraduateSophie Chiang.

Their work started life as a project during a student module last autumn on Affective AI. This is course on a branch of AI aimed at creating artificially intelligent systems and machines that can recognise, interpret, process, and simulate, human social signals and behaviours, expressions, and emotions.

As part of the project, Tom and Sophie were looking at 'vision-language' AI models, the development of which is advancing rapidly. These are multi-modal models that can analyse and interpret data not only in text, but also in audio, and still and moving images.

The project idea was originally proposed by Dr Jiaee Cheong and Dr Irmak Dogan, postdoctoral researchers whose expertise spans fairness, explainability, and mental wellbeing assessment.

There is growing interest in using such vision-language models as potential tools to help clinicians assess patients' mental health. "But their deployment in clinical settings has raised concerns due to their lack of transparency and potential for bias," say the two students in their paper.

Because of the 'black box' nature of the models, they explain, it can make it "difficult for clinicians to discern whether a classification is based on valid clinical markers or a result of machine learning prediction bias".

Tom and Sophie were interested in finding out if such models could interpret recordings of patients who were being assessed for depression both fairly and accurately, and yet still do so explainably.

So, they set out to explore how two models' diagnostic reliability and demographic fairness fared when frameworks were applied to improve the explainability of the decisions they reached.

They were looking at datasets of interview with patients in two different settings; in one, patients were in a one-to-one session with a coach, in another the setting was more clinical. And they found that "performance varied substantially across environments and architectures".

While there was bias in both models, one showed more gender bias, they said, while another exhibited more racial bias. They added that their results highlight a persistent gap between procedural transparency and equitable outcomes.

Tom and Sophie sent in their end-of-module report for marking last January. In return, they received in return an email encouraging them to consider extending their work into a conference paper. With some additional support and mentoring from members of the Department, they did so and submitted it to the ACM International Conference On Multimodal Interaction.

In June, their paper was accepted – and this week Tom Brennan is in Naples to attend the conference and present their findings.

This news has delighted their Affective AI lecturer Hatice Gunes, who is Professor of Affective Intelligence and Robotics here and who, together with Drs Dogan and Cheong, supported and mentored Tom and Sophie in developing the project into a full conference paper.

"This is a particularly remarkable achievement at this stage of Tom's and Sophie's studies," she says. "The acceptance rate for full papers at this conference is typically below 36% this year. This highlights the high-quality research our students can contribute to, even at the undergraduate level."

Dr Irmak Dogan adds, "It was a great pleasure to see how Sophie and Tom engaged with the research questions around fairness and explainability, grasped new concepts, and incorporated feedback effectively, leading to the work being accepted at a major international conference."

And Dr Jiaee Cheong says "I am incredibly proud of what they achieved and of the hard work, dedication, and perseverance showed throughout."

Both students say that the experience of taking part in research has whetted their appetite for more.

Sophie has already embarked on the optional fourth year of the Computer Science degree here, which is an integrated master's course, designed to support students who are considering a career in academic or industrial research. She is thinking about following it up with a PhD.

As for Tom,"it's definitely made me want to get more involved in research and see where else I can contribute," he says. "It's been a pretty fun experience and very rewarding."