ChatThero: An LLM-Supported Chatbot for Behavior Change and Therapeutic Support in Addiction Recovery

Substance use disorders (SUDs) affect over 36 million people worldwide, yet only a small fraction seek or receive effective treatment due to stigma, motivational barriers, and insufficient personalized support. Although large language models (LLMs) have shown significant potential in mental health support applications, current systems often lack deep integration of clinically validated therapeutic strategies, limiting their efficacy in challenging contexts such as addiction recovery.In this work, we introduce ChatThero, an innovative multi-agent conversational framework specifically designed for scalable, long-term addiction treatment support. ChatThero uniquely integrates dynamic patient modeling, context-sensitive therapeutic dialogues, and adaptive persuasive strategies grounded in cognitive behavioral therapy (CBT) and motivational interviewing (MI). We construct a high-fidelity synthetic benchmark comprising realistic recovery scenarios categorized by patient resistance levels (Easy, Medium, and Hard), and employ a two-stage training approach combining supervised fine-tuning (SFT) with direct preference optimization (DPO).Comprehensive evaluation demonstrates ChatThero’s substantial improvements over state-of-the-art models, achieving a 41.5% average gain in patient motivation, an increase of 0.49% in treatment confidence, and resolving challenging (hard-level) cases with 26% fewer conversational turns compared to GPT-4o. Moreover, automated clinical evaluations show that ChatThero significantly surpasses GPT-4o and other leading LLMs in empathy, responsiveness, and behavioral realism. Our framework not only facilitates rigorous and ethically responsible exploration of conversational strategies in addiction care but also provides a robust, replicable foundation for the immediate practical deployment of AI-driven therapeutic interventions in real-world community and clinical settings.