Conversational Self-Play for Discovering and Understanding Psychotherapy Approaches

2025-03-17

AI Summary

This work addresses the challenges of automatic identification and interpretable modeling of psychotherapeutic orientations. We propose a dialogue-based self-play framework leveraging large language models (LLMs). Methodologically, we design a dual-role, therapy-aware dialogue mechanism that integrates prompt engineering with multidimensional evaluation metrics—namely theoretical consistency, technical behavior coverage, and clinical plausibility—to enable unsupervised deconstruction and alignment of therapeutic approaches. Innovatively, we introduce self-play to psychotherapy methodology research, enabling, for the first time, automated discovery and validation of core intervention patterns across distinct therapeutic orientations. Experiments successfully reconstruct and identify hallmark dialogue patterns for seven major evidence-based therapies—including CBT and ACT—with an expert-blind evaluation accuracy of 86.4% in modality identification. Furthermore, our analysis uncovers shared intervention logic structures across otherwise divergent therapeutic schools.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

This paper explores conversational self-play with LLMs as a scalable approach for analyzing and exploring psychotherapy approaches, evaluating how well AI-generated therapeutic dialogues align with established modalities.

Problem

• Analyzing psychotherapy approaches via conversational self-play
• Evaluating AI-generated therapeutic dialogue alignment
• Exploring scalable methods for psychotherapy understanding

Innovation

• LLMs enable scalable psychotherapy analysis
• Conversational self-play explores therapeutic dialogues
• AI-generated dialogues align with modalities