arXiv cs.CY· Jiwon Kim, Violeta J. Rodriguez, Dong Whi Yoo, Eshwar Chandrasekharan, Koustuv Saha·· 12 小时前AI 评分34
PAIR-SAFE：用配对智能体在运行时审计并改进 AI 心理健康支持
PAIR-SAFE: A Paired-Agent Approach for Runtime Auditing and Refining AI-Mediated Mental Health Support
AI 导读
PAIR-SAFE 是一个配对智能体框架，将 Responder 智能体与基于临床验证的 MITI-4 的 Judge 智能体结合，由 Judge 对每条回复给出 allow 或 revise 决策以实时改进。
正文 · 原文
Abstract:LLMs are increasingly used for mental health support, yet they can produce responses that are overly directive, inconsistent, or clinically misaligned. Most risk mitigation approaches rely on implicit alignment through training or prompting, offering limited runtime accountability. We introduce PAIR-SAFE, a paired-agent framework that integrates a Responder agent with a Judge agent grounded in the clinically validated Motivational Interviewing Treatment Integrity (MITI-4). The Judge audits each response and issues \allow{} or \revise{} decisions to guide real-time refinement. We evaluate interactions using a support-seeker simulator derived from human-annotated data. The Judge-supervised interactions show significant improvements in key MITI dimensions, including Partnership, Seek Collaboration, and overall Relational quality. Our quantitative findings are corroborated by expert evaluation, highlighting nuances of runtime supervision. Our study reveals that paired-agent auditing can provide clinically grounded, transparent oversight for AI-assisted mental health support.
| Subjects: | Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY) |
| Cite as: | arXiv:2601.12754 [cs.HC] |
| (or arXiv:2601.12754v2 [cs.HC] for this version) | |
| https://doi.org/10.48550/arXiv.2601.12754 arXiv-issued DOI via DataCite |
Submission history
From: Koustuv Saha [view email]
[v1]
Mon, 19 Jan 2026 06:20:57 UTC (1,105 KB)
[v2]
Fri, 23 Oct 2026 15:04:11 UTC (10,110 KB)
来源：arXiv cs.CY · arxiv.org