Tweaking the Messages and Approaching the Glass Box: Using AI Chatbots to Promote Help-Seeking for Depressive Symptoms

Abstract

This study tests how AI-based health chatbots’ message framing, along with the explanations about human knowledge involvement in their algorithms, influence users’ attitudes toward chatbots’ recommendation. Based on a two-level human-machine communication framework, an online experiment (N = 374) revealed that a chatbot’s explanations of the high (vs. low) involvement of human knowledge in its algorithms increased users’ trust in the chatbot, which further improved their attitudes toward help-seeking. The message target (targeted vs. mistargeted) employed in the chatbot’s recommendations, the involvement of human knowledge in the algorithms, and users’ depression tendency jointly influenced users’ psychological reactance, which further affected their attitudes toward seeking help from friends and family members. Our findings can contribute to current understandings of how AI shapes the persuasive mechanism of health promotion messages and offer insights into using AI for mental health promotion.