Beyond Compliance: A Resistance-Informed Motivation Reasoning Framework for Challenging Psychological Client Simulation

arXiv cs.AI / 4/14/2026

📰 NewsSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces ResistClient, a psychological client simulator designed to overcome limitations of existing models that tend to produce unrealistic over-compliance during counseling training and evaluation.
It grounds simulated “challenging client behaviors” in Client Resistance Theory and uses a two-stage framework called Resistance-Informed Motivation Reasoning (RIMR) to connect external behaviors with internal motivational mechanisms.
RIMR first reduces compliance bias through supervised fine-tuning on RPC, a large resistance-oriented psychological conversation dataset spanning diverse client profiles.
It then goes beyond response imitation by training models to produce psychologically coherent motivation reasoning before generating responses, with joint optimization for motivation authenticity and response consistency using process-supervised reinforcement learning.
The authors report extensive automatic and expert evaluations showing improved challenge fidelity, behavioral plausibility, and reasoning coherence, and highlight potential for better evaluation and optimization of mental-health dialogue LLMs under difficult scenarios.

Abstract

Psychological client simulators have emerged as a scalable solution for training and evaluating counselor trainees and psychological LLMs. Yet existing simulators exhibit unrealistic over-compliance, leaving counselors underprepared for the challenging behaviors common in real-world practice. To bridge this gap, we present ResistClient, which systematically models challenging client behaviors grounded in Client Resistance Theory by integrating external behaviors with underlying motivational mechanisms. To this end, we propose Resistance-Informed Motivation Reasoning (RIMR), a two-stage training framework. First, RIMR mitigates compliance bias via supervised fine-tuning on RPC, a large-scale resistance-oriented psychological conversation dataset covering diverse client profiles. Second, beyond surface-level response imitation, RIMR models psychologically coherent motivation reasoning before response generation, jointly optimizing motivation authenticity and response consistency via process-supervised reinforcement learning. Extensive automatic and expert evaluations show that ResistClient substantially outperforms existing simulators in challenge fidelity, behavioral plausibility, and reasoning coherence. Moreover, ResistClient facilities evaluation of psychological LLMs under challenging conditions, offering new optimization directions for mental health dialogue systems.