![[2602.22775] TherapyProbe: Generating Design Knowledge for Relational Safety in Mental Health Chatbots Through Adversarial Simulation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.22775] TherapyProbe: Generating Design Knowledge for Relational Safety in Mental Health Chatbots Through Adversarial Simulation
Summary
The paper introduces TherapyProbe, a methodology for enhancing relational safety in mental health chatbots through adversarial simulations, identifying failure patterns and design recommendations.
Why It Matters
As mental health chatbots become more prevalent, ensuring their effectiveness and safety is crucial. This research addresses the limitations of current evaluation methods by focusing on the dynamics of chatbot interactions over time, providing a framework for safer and more effective mental health support.
Key Takeaways
- TherapyProbe offers a novel methodology for evaluating chatbot interactions.
- Identifies 23 failure archetypes that can harm user experience.
- Provides actionable design recommendations for developers and clinicians.
Computer Science > Human-Computer Interaction arXiv:2602.22775 (cs) [Submitted on 26 Feb 2026] Title:TherapyProbe: Generating Design Knowledge for Relational Safety in Mental Health Chatbots Through Adversarial Simulation Authors:Joydeep Chandra, Satyam Kumar Navneet, Yong Zhang View a PDF of the paper titled TherapyProbe: Generating Design Knowledge for Relational Safety in Mental Health Chatbots Through Adversarial Simulation, by Joydeep Chandra and 2 other authors View PDF HTML (experimental) Abstract:As mental health chatbots proliferate to address the global treatment gap, a critical question emerges: How do we design for relational safety the quality of interaction patterns that unfold across conversations rather than the correctness of individual responses? Current safety evaluations assess single-turn crisis responses, missing the therapeutic dynamics that determine whether chatbots help or harm over time. We introduce TherapyProbe, a design probe methodology that generates actionable design knowledge by systematically exploring chatbot conversation trajectories through adversarial multi-agent simulation. Using open-source models, TherapyProbe surfaces relational safety failures interaction patterns like "validation spirals" where chatbots progressively reinforce hopelessness, or "empathy fatigue" where responses become mechanical over turns. Our contribution is translating these failures into a Safety Pattern Library of 23 failure archetypes with corresponding des...
