ChatThero: An LLM-Supported Chatbot for Behavior Change and Therapeutic Support in Addiction Recovery
This addresses the challenge of recovery support for individuals with addiction by offering a low-cost, multi-session chatbot, though it is incremental as it builds on existing LLM and therapy methods.
The paper tackles the problem of limited access to care for substance use disorders by developing ChatThero, an autonomous language agent that provides therapeutic support, resulting in significant improvements such as raising motivation by +1.71 points and confidence by +1.67 points on a 1-5 scale, and achieving success with difficult patients in 26% fewer turns than GPT-5.
Substance use disorders (SUDs) affect millions of people, and relapses are common, requiring multi-session treatments. Access to care is limited, which contributes to the challenge of recovery support. We present \textbf{ChatThero}, an innovative low-cost, multi-session, stressor-aware, and memory-persistent autonomous \emph{language agent} designed to facilitate long-term behavior change and therapeutic support in addiction recovery. Unlike existing work that mostly finetuned large language models (LLMs) on patient-therapist conversation data, ChatThero was trained in a multi-agent simulated environment that mirrors real therapy. We created anonymized patient profiles from recovery communities (e.g., Reddit). We classify patients as \texttt{easy}, \texttt{medium}, and \texttt{difficult}, three scales representing their resistance to recovery. We created an external environment by introducing stressors (e.g., social determinants of health) to simulate real-world situations. We dynamically inject clinically-grounded therapeutic strategies (motivational interview and cognitive behavioral therapy). Our evaluation, conducted by both human (blinded clinicians) and LLM-as-Judge, shows that ChatThero is superior in empathy and clinical relevance. We show that stressor simulation improves robustness of ChatThero. Explicit stressors increase relapse-like setbacks, matching real-world patterns. We evaluate ChatThero with behavioral change metrics. On a 1--5 scale, ChatThero raises \texttt{motivation} by $+1.71$ points (from $2.39$ to $4.10$) and \texttt{confidence} by $+1.67$ points (from $1.52$ to $3.19$), substantially outperforming GPT-5. On \texttt{difficult} patients, ChatThero reaches the success milestone with $26\%$ fewer turns than GPT-5.