New arXiv study: Empathetic AI chatbots reshape health communication
48-person experiment shows empathy and context change how users share health info
A new study from the Human-Computer Interaction lab, led by Xi Zheng and three co-authors, examines how empathetic expression and conversational context influence user engagement in AI-powered health inquiries. Published on arXiv (2608.17144) and set to appear at UbiComp Companion 2026, the researchers ran a 2x2x3 within-subjects experiment with 48 participants, testing chatbot empathy modalities—verbal, visual, and multimodal—across general, sensitive, and mental health contexts.
The results showed that verbal and multimodal empathy significantly increased reply length, but communicative acts were primarily driven by context. Sensitive contexts triggered more question-asking, while mental health contexts led to heightened concerns, assertive responses, and unprompted information disclosure. Qualitative insights point to design implications for building context-sensitive AI health systems that actively encourage user participation. This work underscores that empathy alone isn't enough—AI health chatbots must adapt to emotional and situational nuances.
- Verbal and multimodal empathy increased user reply length in health AI conversations
- Sensitive contexts triggered 2x more question-asking behavior from users
- Mental health contexts led to assertive responses and unprompted info disclosure
Why It Matters
Context-aware AI health chatbots can improve patient engagement and data sharing for better diagnostics.