生成式人工智能赋能高校心理健康服务的兴起、挑战与规范化发展路径

The Rise, Challenges, and Path to Standardized Development of Generative AI-Empowered Mental Health Services in Higher Education

摘要: 高校学生心理健康问题日益严峻，传统心理健康服务体系在专业资源、服务可及性、学生病耻感等方面面临多重困境。生成式人工智能(Generative Artificial Intelligence, GenAI)技术的迅猛发展为破解上述难题提供了新的技术路径。现有研究表明，人工智能聊天机器人在心理健康科普、情绪支持、自我调节和部分数字化干预任务中具有扩大服务可及性、降低使用门槛和增强个性化交互的潜力，但其实际应用仍面临专业边界模糊、数据隐私、风险识别、情感依赖及责任界定难等问题。本文基于国内外相关研究，对生成式人工智能赋能高校心理健康服务的兴起、技术演进和应用现状进行梳理，并进一步讨论提示工程与指令微调、检索增强生成(RAG)、情感计算和多角色协作架构等关键技术机制。在此基础上从功能分层、专业知识约束、安全评估、数据治理、人机协同和校园生态融合以及中国高校文化适应等方面提出规范化发展路径。研究认为，高校生成式人工智能心理健康服务的定位不应替代专业人员，而是在以人为中心和专业人员监督的前提下，承担低风险、高频次、即时性的心理支持与情绪疏解任务，并通过有效转介机制与高校现有心理健康服务体系形成互补，从而实现技术效率与心理服务专业性的平衡。

Abstract: Mental health issues among university students are becoming increasingly severe, and traditional mental health service systems face multiple challenges regarding professional resources, service accessibility, and student stigma. The rapid development of Generative Artificial Intelligence (GenAI) offers a new technological pathway to address these difficulties. Existing research indicates that AI chat bots have the potential to expand service accessibility, lower barriers to use, and enhance personalized interaction in areas such as mental health education, emotional support, self-regulation, and certain digital intervention tasks; however, their practical application faces issues such as blurred professional boundaries, data privacy concerns, risk identification challenges, emotional dependency, and difficulties in defining liability. Drawing on relevant domestic and international research, this paper reviews the emergence, technological evolution, and current application status of GenAI-enabled mental health services in universities. It further discusses key technical mechanisms, including prompt engineering and instruction fine-tuning, Retrieval-Augmented Generation (RAG), affective computing, and multi-role collaborative architectures. Based on this analysis, the paper proposes a pathway for standardized development, covering aspects such as functional stratification, professional knowledge constraints, safety assessment, data governance, human-AI collaboration, integration into the campus ecosystem, and cultural adaptation within Chinese universities. The study concludes that GenAI-based mental health services in universities should not replace professionals; instead, operating under a human-centric approach and professional supervision, they should undertake low-risk, high-frequency, and immediate tasks related to psychological support and emotional relief. By establishing effective referral mechanisms, these services can complement existing university mental health systems, thereby achieving a balance between technological efficiency and the professionalism of mental health services.