Bridging Human-Computer Interaction and Behavioural Science for Fair and Trustworthy Large Language Model ‑ Driven Financial Advice
Bridging Human-Computer Interaction and Behavioural Science for Fair and Trustworthy AI in Personal Finance
In partnership with the Financial Conduct Authority, this BR-UK funded research project aims to produce actionable insights that support the responsible use of LLMs in consumer finance, making high-quality financial advice more accessible and trustworthy
Project team
- Dr Yulu Pi, Principal Investigator, Centre for Interdisciplinary Methodologies, University of Warwick
- Prof. Cagatay Turkay, Co-Investigator, Centre for Interdisciplinary Methodologies, University of Warwick
- Daniel Bogiatzis-Gibbons, Collaborator, Financial Conduct Authority
- Prof. Jatinder Singh, Collaborator, University of Cambridge
- Jinha Yoon, Research Assistant, University of Warwick
Contact
For more information, contact Dr Yulu Pi at yulu.pi@warwick.ac.uk.
This project investigates how people understand, trust, and act on financial advice from AI chatbots. In partnership with the Financial Conduct Authority (FCA), it turns the findings into design and policy guidance for responsible AI in personal finance. It is funded by a
Behavioural Research UK (BR-UK) Commissioning Fund Accelerator Award.
About the project
Millions of people in the UK struggle to access affordable, reliable financial advice, a problem known as the "financial advice gap", highlighted by the UK Government and the Financial Conduct Authority (FCA). New technologies like Large Language Models (LLMs) hold the promise to provide low-cost, scalable, and accessible financial guidance, particularly for those underserved by traditional advice channels. However, fulfilling this promise requires more than just LLMs' technological capability; it hinges on how consumers understand, trust, and act on advice these systems give.
To tackle this challenge, this interdisciplinary project brings together experts in Human–Computer Interaction (HCI), behavioural science, and responsible AI. We start by identifying the risks and limitations of using LLMs for personal finance advice through focus groups that explore users' initial attitudes and concerns. Building on these insights, we will conduct experiments to test how different interaction features of LLM-generated advice, such as the level of personalisation, the tone of delivery (for example, cautious versus confident), and the explanations about how advice is reached, impact users' financial awareness, trust, and decision-making, with a specific focus on potential issues around fairness, trustworthiness, and uninformed use.
Our goal is to produce actionable insights that support the responsible use of LLMs in consumer finance, making high-quality financial advice more accessible and trustworthy. In partnership with the FCA, we aim to translate the behavioural findings gained from this project into practical design and policy guidance to inform the real-world deployment of LLM-driven financial tools.
Research question
How can AI in personal finance, from AI chatbots that give financial advice to AI agents that act on people's behalf, be designed to promote understanding, appropriate trust, and informed use?