AI Conversational Agents in Healthcare for Type-2 Diabetes

Abstract

Type 2 diabetes (T2D) is a global health crisis with significant impacts on individuals and healthcare systems. This thesis develops AI conversational agents (CAs) to promote physical activity and lifestyle changes for those at risk of T2D through a multi-phase study, including a systematic review, a design framework, and empirical testing.

The systematic review identified gaps in digital interventions, particularly the limited use of CAs in T2D prevention. A standardised framework was then developed, focusing on personalisation, user engagement, and proactive health management. This framework guided the iterative design and refinement of a CA prototype, tested across diverse populations in Sydney and Jeddah. The thesis integrated real-time activity tracking via Fitbit and enhanced conversational capabilities using large language models. Findings demonstrated that AI-driven, personalised interactions significantly encouraged physical activity, a key factor in preventing T2D progression.

This thesis contributes to health informatics by demonstrating AI’s role in preventive healthcare. It highlights the importance of a user-centred design approach, ensuring that digital health tools are effective and align with the users’ needs and preferences. Future research should focus on long-term engagement strategies and integrating conversational agents with broader healthcare systems to enhance their effectiveness and reach.