A Framework for ECA-Based Psychotherapy

Abstract

We present Tessa, an autonomous framework for structured bi-weekly psychotherapy intervention. Tessa alternates between screening sessions that use motivational interviewing and follow-up CBT sessions to help users identify, challenge, and reframe negative thoughts. Using text analysis with Large Language Models (LLMs), an Embodied Conversational Agent (ECA) adapts its dialogue and synchronizes non-verbal cues in real time. The framework incorporates a therapist-matching pipeline and clinical report generation to enable seamless hand-off to professional care. Taken together, Tessa provides accessible, stigma-free, and data-driven mental health support that improves personalized care and reduces barriers to therapy.