Jason Chan

I am a third-year PhD student at the University of Sheffield, co-supervised by Prof. Rob Gaizauskas and Dr. Cass Zhixue Zhao. My research evaluates and improves the reasoning capabilities of large language models (LLMs) by drawing on cognitive science theories of human reasoning.

In particular, I highlight how an over-reliance on formal logic undermines LLMs’ ability to reason with natural language and creates vulnerabilities in downstream applications such as fact-checking. To address this, I am developing methods inspired by the mental model theory that enable LLMs to reason flexibly through explanations and counterexamples. In doing so, I aim to improve how these models interact with humans and how they perform in complex domains such as science and law, where formulating the right hypotheses and questions is just as important as getting to the right answers.

Prior to my PhD, I completed an MSc in Speech and Language Processing at the University of Edinburgh, and a BA in Philosophy at UCL. I also hold legal qualifications and have previously worked as a financial services regulatory lawyer in a City law firm in the UK.

I am open to internships, research collaborations and future industry roles, particularly in the UK and North Rhine-Westphalia, Germany.