SimCounsel: Emotion-Aware Simulation of Multi-Agent Counseling

ACL ARR 2025 May Submission4700 Authors

20 May 2025 (modified: 03 Jul 2025)ACL ARR 2025 May SubmissionEveryoneRevisionsBibTeXCC BY 4.0
Abstract: While counseling is a promising domain for applying large language model (LLM) based multi-agent simulations, research in this area remains in its early stages. In this study, we propose SimCounsel, a multi-agent simulation framework designed to explore the effectiveness of counseling interventions. SimCounsel incorporates emotion-aware memory retrieval mechanisms and persona update strategies based on real psychological theories, enabling more human-like and context-sensitive agent behavior. Both automated evaluations and expert evaluations confirmed that the proposed framework not only realistically simulates the behavior of real clients and therapists, but also effectively facilitates analysis of counseling outcomes. These findings suggest that SimCounsel holds significant promise as a scalable and research-friendly tool for counseling-related simulations. All the code for the simulation is presented on GitHub.
Paper Type: Long
Research Area: Computational Social Science and Cultural Analytics
Research Area Keywords: human behavior analysis, NLP tools for social analysis
Contribution Types: NLP engineering experiment
Languages Studied: English
Submission Number: 4700
Loading