Encouraging Deep Self-Disclosure in AI-Enabled Mental Health Support Systems through Politeness Strategies

AI-enabled Mental Health Support Systems (AIMHSS) are increasingly used to expand access to mental health support. Yet effective support requires more than empathy and validation: during cognitive restructuring, these systems may need to challenge maladaptive beliefs. Such feedback may cause value conflicts and be perceived as a face-threatening act (FTA), weakening the deep self-disclosure needed for support. Prior human-AI interaction research has examined supportive socio-emotional cues, but offers limited guidance on how AI should deliver challenging feedback without discouraging disclosure. Drawing on Politeness Theory, we propose that positive and negative face strategies may reduce perceived self-threat and thereby promote deep self-disclosure intention, and that AIMHSS role configuration (expert vs. partner) moderates these effects. We plan to conduct a scenario-based experiment to empirically test the proposed research model. This research is expected to guide the design of AIMHSS that deliver necessary challenges while preserving users’ willingness to disclose.