July 23, 2025

Transformer-based Large Language Models (LLMs) have significantly accelerated the development of AI-assisted applications in real-world settings. In the business operation, balancing the need for robust, data-driven AI models with the ethical imperatives of fairness and privacy is a complex, yet essential, task.

In response to these challenges, there is a growing recognition that Generative AI (GenAI) models must be designed with a broader set of criteria in mind. There is also growing attention in the research community toward aligning AI models with real-world societal needs and values. Although various solutions have been proposed to address these concerns, there remains limited research on their effectiveness within the insurance sector. This project aims to bridge the gap between emerging AI fairness evaluation methods, debiasing techniques, and model interpretability, proposed in computer science or statistics communities—and the current lack of structured ethical and governance frameworks for practical AI implementation in the insurance sector. We try to identify AI solutions that are appropriate for the insurance sector and translate them into actuarial terms to facilitate easier understanding and adoption within the profession.

This project aims to address the following key areas and develop practical guidelines to support actuarial practices in the ethical and responsible use of GenAI systems within the insurance sector: (1) Deploying and Operationalizing GenAI for Insurance Use Cases; (2) Ensuring Reliability and Governance in GenAI; (3) Fair and Ethical GenAI; (4) GenAI Compliance with Insurance Regulatory.

Supervisor: Zhiyu (Frank) Quan
Graduate Supervisor: Panyi Dong