How Explainability Shapes User Perception and Therapeutic Effectiveness in Mental Healthcare Chatbots: Accountability as a Boundary Condition

Mental healthcare chatbots have emerged as an accessible therapeutic resource. Owing to the sensitivity of its usage scenario, an appropriate design is essential to ensure a secure and comfortable user experience, with particular emphasis on explainability and accountability. However, how explainability and accountability jointly influence use experience and therapeutic effectiveness remains ambiguous. This research aims to explore the optimal design of mental healthcare chatbots that balance explainability and accountability via an experimental design. Drawing on feelings-as-information theory, this study examines how explainability influences users' perceived empathy, perceived expertise, and processing fluency, and how these effects vary across accountability levels. The findings of this study extend the feelings-as-information theory to human-AI interaction in the context of mental healthcare, revealing how these mechanisms operate under varying accountability levels, while offering actionable design guidelines for developing explainable and accountable mental healthcare chatbots.