Reproduce “Bloom: Designing for LLM-Augmented Behavior Change Interactions” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Matthew Jörke, Defne Genç, Valentin Teutschbein, Shardul Sapkota, Sarah Chung, Paul Schmiedmayer, Maria Ines Campero, Abby C. King and 2 more
Bloom, a mobile application, integrates a large language model (LLM)-based health coaching chatbot with established multimodal behavior change interactions to promote physical activity. A randomized field study revealed that the LLM-augmented version fostered more positive mindsets, increased app engagement by 5.6 times daily, and facilitated the creation of more personalized activity plans compared to a no-LLM control.
Key insights
- LLMs can augment established multimodal HCI behavior change interactions by interpreting qualitative user context to personalize UI elements and interactions, moving beyond text-only coaching.
- An empathetic and non-judgmental conversational style from the LLM fosters positive shifts in users' physical activity mindsets, including increased enjoyment, self-confidence, and self-compassion, promoting user agency.
- Relational cues from the AI coach, even when its AI nature is known, significantly increase user engagement across all application features, not just the chat interface.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.