Mixed Chain-of-Psychotherapies for Emotional Support Chatbot

Siyuan Chen,Cong Ming,Zhiling Zhang,Yanyi Chen,Kenny Q. Zhu,Mengyue Wu

2024-09-29

Abstract:In the realm of mental health support chatbots, it is vital to show empathy and encourage self-exploration to provide tailored solutions. However, current approaches tend to provide general insights or solutions without fully understanding the help-seeker's situation. Therefore, we propose PsyMix, a chatbot that integrates the analyses of the seeker's state from the perspective of a psychotherapy approach (Chain-of-Psychotherapies, CoP) before generating the response, and learns to incorporate the strength of various psychotherapies by fine-tuning on a mixture of CoPs. Through comprehensive evaluation, we found that PsyMix can outperform the ChatGPT baseline, and demonstrate a comparable level of empathy in its responses to that of human counselors.

Computation and Language

What problem does this paper attempt to address?

The paper aims to address two main issues of emotional support chatbots in psychological counseling: "surface empathy" and "quick fixes." Specifically:

1. **Surface Empathy**: Current methods often provide vague and generalized responses, lacking a deep understanding of the user's context, making users feel inadequately understood.
2. **Quick Fixes**: Large language models (LLMs) tend to offer general solutions immediately when users express concerns, without exploring underlying causes or less obvious issues.

To address these problems, the research team proposed a chatbot system named PsyMix. This system analyzes the seeker's situation by integrating multiple psychotherapy approaches (such as cognitive-behavioral therapy, person-centered therapy, and solution-focused brief therapy) and generates responses based on this analysis. In this way, PsyMix can demonstrate higher professionalism and specificity in understanding and responding to users, thereby improving the quality of emotional support. Experimental results show that PsyMix exhibits better empathy and conversation quality in human evaluations compared to baseline methods like ChatGPT.