captAIn
Context-Aware artificial intelligence for Personalised just-in-Time interventions in physical ActIvity motivatioN
Physical inactivity is a key contributor to various chronic health issues and diminished overall well-being. Just-in-Time Adaptive Interventions (JITAIs) use real-time data to deliver personalised, timely prompts that support behaviour change when it is most impactful. This dissertation explores how the integration of context-aware computing (e.g., mood, location, stress levels, and social environment) with advanced AI techniques can enhance the design of JITAIs to promote long-term engagement and increases in physical activity.
Research Goals
- Design adaptive models that respond to individual psychological characteristics and real-time contextual factors
- Assess hybrid AI techniques for predicting the most effective behaviour change strategies
- Implement and evaluate an AI-driven intervention system in real-world settings
Project Summaries
- Theory of Mind & Personality Traits
Investigating whether tailoring interventions to Big Five personality traits can enhance users’ perception of personalisation and empathy in JITAIs. - Contextual Multi-Armed Bandit (cMAB) + LLMs
This project combines contextual Multi-Armed Bandits with Large Language Models to select appropriate Behaviour Change Techniques based on current user context and generate personalised motivational messages. The study is conducted in collaboration with the Department of Psychology and Statistical Sciences at the University of Toronto. - AI Pipeline & “Into the Wild” Study
A real-world deployment to study how user feedback contributes to JITAI personalisation. Examining the motivational impact of both pre-planned and spontaneous physical activity, as well as identifying which contextual features different AI models prioritise in the personalisation process.
Anticipated Contributions
- Insights into which contextual signals effectively support physical activity engagement
- An open-source AI-based framework to facilitate future research and development of personalised interventions
Collaborators
- Prof. Dr. Albrecht Schmidt, Ludwig Maximilians University of Munich (LMU)
- Dr.-Ing. Jan David Smeddinck, Ludwig Boltzmann Institute for Digital Health and Prevention (LBI-DHP)
- Dr. David Haag, LBI-DHP
- Dr. Rania Islambouli, LBI-DHP
- Fred Haochen Song, University of Toronto, Canada
Contact Person
