[Paper Review] PsyCLIENT: Client Simulation via Conversational Trajectory Modeling for Trainee Practice and Model Evaluation in Mental Health Counseling

Huachuan Qiu, Zhaoming Chen+4 more|arXiv (Cornell University)|Jan 12, 2026

Digital Mental Health Interventions0 citations

TL;DR

PsyCLIENT introduces a conversational trajectory modeling framework to simulate diverse, authentic Chinese-language client interactions for trainee practice and evaluating LLM-based counselors, achieving high realism and training effectiveness. It outperforms baselines and yields dialogues indistinguishable from human clients.

ABSTRACT

LLM-based client simulation has emerged as a promising tool for training novice counselors and evaluating automated counseling systems. However, existing client simulation approaches face three key challenges: (1) limited diversity and realism in client profiles, (2) the lack of a principled framework for modeling realistic client behaviors, and (3) a scarcity in Chinese-language settings. To address these limitations, we propose PsyCLIENT, a novel simulation framework grounded in conversational trajectory modeling. By conditioning LLM generation on predefined real-world trajectories that incorporate explicit behavior labels and content constraints, our approach ensures diverse and realistic interactions. We further introduce PsyCLIENT-CP, the first open-source Chinese client profile dataset, covering 60 distinct counseling topics. Comprehensive evaluations involving licensed professional counselors demonstrate that PsyCLIENT significantly outperforms baselines in terms of authenticity and training effectiveness. Notably, the simulated clients are nearly indistinguishable from human clients, achieving an about 95\% expert confusion rate in discrimination tasks. These findings indicate that conversational trajectory modeling effectively bridges the gap between theoretical client profiles and dynamic, realistic simulations, offering a robust solution for mental health education and research. Code and data will be released to facilitate future research in mental health counseling.

Motivation & Objective

• Address the lack of diverse and realistic client profiles in LLM-based counseling simulations.
• Propose a principled, trajectory-based framework to model realistic client behaviors.
• Create PsyCLIENT-CP, the first open-source Chinese client profile dataset covering 60 topics.
• Demonstrate through expert evaluations that trajectory-based simulation improves authenticity and training effectiveness.

Proposed method

• Formalize client behavior as a discrete, multi-label space and define behavior trajectories.
• Define counselor strategy space and a mapping from client behaviors to valid counselor responses.
• Develop G to generate utterances conditioned on client profiles and target behavior sets, constrained to linguistic realizations L_B.
• Use a prompt design that incorporates client profile, dialogue history, behavior labels, and content constraints for LLM-based client simulation.
• Extract conversational trajectories from real dialogues with behavior annotations to guide simulation.
• Open-source PsyCLIENT-CP with 60 topics and 38,880 potential client instances (120 profiles × 324 trajectories).

Experimental results

Research questions

• RQ1 RQ1 Authenticity: Do PsyCLIENT dialogues achieve higher authenticity than baselines according to expert evaluation?
• RQ2 RQ2 Discrimination: Can experts distinguish PsyCLIENT dialogues from human-human or baseline-generated dialogues?
• RQ3 RQ3 Effectiveness: Do counselors rate PsyCLIENT as more effective for training compared to other methods?

Key findings

• PsyCLIENT achieves higher authenticity across fluency, emotion expression, coherence, appropriateness, and overall authenticity than vanilla, content, and behavior baselines.
• Behavior labeling within trajectories yields the largest gains in realism, with content providing additional but smaller benefits.
• Experts nearly cannot distinguish PsyCLIENT dialogues from human-client dialogues (about 95% confusion rate).
• PsyCLIENT outperforms baselines in counseling effectiveness across listening, questioning, emotion-handling, technique-practice, and recommendations.
• Discrimination accuracy for PsyCLIENT is at or near chance for expert judges, indicating high realism in generated dialogues.