Reference and Solution Architecture for GenAI- and GIS-Enhanced Physical Activity Interventions: Towards Implementing...

Area of Science:

• Digital Health
• Human-Computer Interaction
• Behavioral Science

Background:

• Digital Behaviour Change Interventions (DBCIs) utilize Information and Communication Technology (ICT) to improve health.
• Participant fatigue and reduced adherence are significant challenges in DBCIs due to frequent or irrelevant interventions.
• Novel engagement mechanisms are crucial for sustained effectiveness of digital health tools.

Purpose of the Study:

• To present a reference and solution architecture for enhancing DBCIs.
• To integrate Large Language Models (LLMs) and Geographic Information Systems (GIS) for improved context-awareness and personalization.
• To address intervention fatigue and enhance user motivation and adherence in digital health interventions.

Main Methods:

• Developed a reference architecture using open-source technologies and Open APIs.
• Integrated a Large Language Model (LLM) component into the DBCI framework.
• Enhanced LLM integration with a Geographic Information Systems (GIS) element for context-awareness.
• Implemented the AI4Motion platform for pilot testing.

Main Results:

• The AI4Motion platform demonstrated the integration of LLM and GIS for DBCIs.
• The architecture supports personalization and contextualization of digital health interventions.
• The approach addresses key challenges related to user fatigue and adherence in digital health.

Conclusions:

• LLM and GIS integration offers a promising approach to create more adaptive and engaging DBCIs.
• The proposed architecture provides a foundation for developing advanced digital health platforms.
• This work contributes to system design patterns for LLM/GIS-enabled digital platforms supporting Ecological Momentary Assessment (EMA), Experience Sampling Method (ESM), and Just-in-Time Adaptive Interventions (JITAIs).