Empirical Modeling of Therapist-Client Dynamics in Psychotherapy Using LLM-Based Assessments

Researchers used large language models to analyze nearly 2,000 hours of psychotherapy transcripts, creating automated assessments for therapist empathy, client disclosure, and rapport. Their LLM-based measures showed strong alignment with human ratings (mean Pearson r = .66). The analysis revealed therapist empathy and exploration directly increase client disclosure, while rapport helps reduce internal distress. This demonstrates a scalable computational method to model core therapeutic dynamics, offering new tools for understanding and improving therapist training.