【摘要】 对人工智能（AI）教育应用风险产生及扩散机制的清晰认知，是对风险进行全面管控的前提。根据社会技术系统理论，风险并非产生于AI技术、行动者、教育任务或治理结构等单一要素，而在于四类要素之间关系的失衡。据此，提出AI教育应用中风险产生及扩散机制的分析框架。其中，引发个体风险的失衡包括师生信任与AI技术功能之间的失衡、师生调适与任务的智能化重构之间的失衡、师生动机与治理结构的绩效倾向之间的失衡；引发系统风险的失衡包括治理结构支持的成效与AI技术系统投入之间的失衡、治理结构的调整与教育任务重构之间的失衡、治理结构保障与“技术—行动者—任务”协同之间的失衡。个体风险与系统风险之间会相互扩散。基于此，对全国459所高职院校管理者、59 414位教师和713 721位学生的调查数据进行分析，结果表明，学校在推进AI应用过程中普遍存在关系失衡现象，由此带来的个体风险包括对AI生成内容的过度信任削弱学生独立思考能力，甚至助长学术不端；学校“试点示范”类绩效导向难以激发教师持续投入的内驱力，阻碍AI应用走向深入和常态化。系统风险包括AI应用陷入“高投入、低产出”困境；AI技术、师生AI素养、AI任务在学校难以协同推进，不仅削弱AI应用的可持续性，也将加剧学校、地区之间的“AI教育鸿沟”等。更多还原

【Abstract】 A clear understanding of the risk generation and diffusion mechanisms for AI application in education is a prerequisite for comprehensive risk management. According to the sociotechnical systems theory, risks do not result from single elements such as AI technology, actors, educational tasks, or governance structures, but rather from imbalanced relationships among these elements. Based on this, this study proposes an analytical framework for the aforesaid mechanisms. In the framework, the imbalances that trigger individual risks include the imbalance between teacher-student trust and the functions of AI technology, the imbalance between teacher-student adaptation and the intelligent restructuring of educational tasks, and the imbalance between teacher-student motivation and the performance orientation of the governance structure. The imbalances that lead to systemic risks include the imbalance between the outcomes resulting from the support from governance structures and the investment in the AI technology system; the imbalance between the adjustment of the governance structure and the restructuring of educational tasks; and the imbalance between the safeguards from the governance structure and the coordination between technology, actors and tasks. Individual risks and systemic risks may diffuse into each other. Based on this framework, the study analyzed survey data from 459 vocational college administers, 59, 414 teachers and 713, 721 students. The findings indicate that vocational colleges generally exhibit relational imbalances during the promotion of AI application. The resulting individual risks include the following: Excessive trust in AI-generated content weakens students’ independent thinking and even induces their academic misconduct; and the performance orientation driven by "pilot demonstration" projects in vocational colleges fails to stimulate teachers’ intrinsic motivation for sustained investment, thus hindering the in-depth and normalized AI application. The systemic risks include the following: AI application falls into the dilemma of "high input but low output"; and AI technology, the AI competence of teachers and students, and AI tasks cannot be advanced in a coordinated manner in vocational colleges, which not only undermines the sustainable application of AI but also exacerbates the "AI education divide" between vocational colleges and regions.更多还原