arXiv cs.CY· Angela Chen, Siwei Jin, Catherine Bao, Canwen Wang, Yi-Chia Wang, Robert E. Kraut, Tongshuang Wu, Haiyi Zhu·· 13 小时前AI 评分31
基于LLM的心理治疗虚拟患者：AVP 用真实治疗转录实现行为自适应
Empirically Grounded Behavioral Adaptivity for LLM-Based Virtual Patients in Psychotherapy Training
AI 导读
研究者提出 Adaptive Virtual Patient（AVP），让基于 LLM 的虚拟患者根据受训者技能在戒备、适度开放到完全披露之间调整自我披露程度。
正文 · 原文
Abstract:Simulated patients offer a scalable way to train psychotherapy micro-skills such as empathic responding and exploratory probing, but current systems either follow fixed scripts or rely on LLMs that drift unpredictably over long sessions. We present the Adaptive Virtual Patient (AVP), which adapts its disclosure behavior in response to trainee skill, from guarded, through moderate openness, to full disclosure. The AVP is informed by a structural equation model fit to nearly 2,000 hours of real-world psychotherapy transcripts, which quantifies how therapist empathy and exploration shift a patient's openness over time. An LLM generates the AVP's utterances conditioned on a disclosure level that the dynamics module updates each turn. In an evaluation with 20 clinicians and trainees over 80 sessions (1,033 turns), the AVP's disclosure rises in response to therapist empathy and exploration, while a prompt-only baseline stays flat; ablations confirm that the empirically motivated parameterization outperforms alternatives, with exploration carrying most of the adaptive signal.
| Comments: | EMNLP 2026 |
| Subjects: | Computers and Society (cs.CY); Human-Computer Interaction (cs.HC) |
| Cite as: | arXiv:2606.10051 [cs.CY] |
| (or arXiv:2606.10051v2 [cs.CY] for this version) | |
| https://doi.org/10.48550/arXiv.2606.10051 arXiv-issued DOI via DataCite |
Submission history
From: Haiyi Zhu [view email]
[v1]
Mon, 8 Jun 2026 18:27:22 UTC (723 KB)
[v2]
Thu, 22 Oct 2026 14:40:54 UTC (1,557 KB)
来源：arXiv cs.CY · arxiv.org