arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.06910cs.CL
收藏

语言塑造大型语言模型中的心理健康评估

Language Shapes Mental Health Evaluations in Large Language Models

University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校);Arizona State University(亚利桑那州立大学)

Jiayi Xu, Xiyang Hu

AI总结:

研究多语言LLM在心理健康评估中是否因语言不同而产生系统性偏差,发现中文提示比英文提示导致更高的污名相关评分和更保守的抑郁严重度判断。

摘要:

多语言大型语言模型(LLMs)越来越多地用于社会敏感的心理健康场景,包括支持聊天机器人、筛查和内容审核。这引发了一个可靠性问题:语义上等效的心理健康输入是否在不同语言中引发可比较的评估,还是会出现与语言相关的社会和文化背景一致的系统性偏移?我们在英中双语环境中使用GPT-4o和Qwen3-32B,通过一个两层框架来检验这个问题:结构层面的评估取向(通过心理测量污名工具测量)和决策层面的行为(通过二元污名检测和四类抑郁严重度分类测量)。在多种工具和模型中,中文提示比英文提示引发更高的污名相关分数。在决策层面,中文提示降低了对污名化内容的敏感性,并产生更保守的抑郁严重度判断,导致更多的低估错误。这些发现表明,提示语言可以改变基于LLM的心理健康评估中的评估取向和下游行为。它们强调了评估多语言LLM时不仅需要关注整体性能,还需要关注它们是否在社会敏感领域中对不同语言应用了可比较的评估标准。

英文摘要:

Multilingual large language models (LLMs) are increasingly used in socially sensitive mental health contexts, including support chatbots, screening, and content moderation. This raises a reliability question: do semantically equivalent mental health inputs elicit comparable evaluations across languages, or systematic shifts consistent with language-associated social and cultural contexts? We examine this question in an English-Chinese setting with GPT-4o and Qwen3-32B using a two-level framework: construct-level evaluative orientation, measured by psychometric stigma instruments, and decision-level behavior, measured by binary stigma detection and four-class depression severity classification. Across instruments and models, Chinese prompts elicit higher stigma-related scores than English prompts. At the decision level, Chinese prompts reduce sensitivity to stigmatizing content and produce more conservative depression severity judgments, leading to more under-estimation errors. These findings show that prompt language can shift both evaluative orientation and downstream behavior in LLM-based mental health evaluation. They highlight the need to evaluate multilingual LLMs not only for aggregate performance, but also for whether they apply comparable evaluative standards across languages in socially sensitive domains.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑