【摘要】 跨学科主题学习作为突破传统学科界限、培育核心素养的关键路径，已成为全球教育改革的重要方向，但在实践中面临教师跨学科主题设计能力不足、学生所获学习支持有限、传统评价体系对素养的捕捉不足等结构性困境。教育多智能体经历了从工具化到智能化、从通用化到场景化、从个体化到协同化的演变过程，具有分布式认知与功能专业化、动态适应性与自组织进化、多模态交互与情境感知能力、协同决策与集体智能涌现等核心特征，并在赋能跨学科主题学习方面展现出了多维度的应用价值，为破解困境提供了有效工具。教育多智能体赋能跨学科主题学习的实现，需要在设计上采用串联式引导驱动、在实施上进行平行式角色协同赋能、在评价上实现反馈式自适应干预。在应用教育多智能体赋能跨学科主题学习时，需依托多智能体协作架构实现系统效能的最大化、应用多智能体打造在地化知识生态与实践场景、设计以学习者为中心的人机协同模式、建立多智能体支持的素养评价与干预闭环、保障多智能体运行的资源基础与伦理规范。更多还原

【Abstract】 Interdisciplinary thematic learning, as a key approach to breaking through traditional disciplinary boundaries and cultivating core competencies, has become an important direction of global educational reform. However, in practice, it faces structural challenges such as insufficient teachers’ capacity to design interdisciplinary themes, limited learning support for students, and the insufficient capture of competencies by traditional assessment systems. Educational multi-agents had undergone an evolution process from instrumentalization to intelligentization, from generalization to scenario-based specialization, and from individualization to collaboration, featuring core characteristics such as distributed cognition and functional specialization, dynamic adaptability and self-organizing evolution, multimodal interaction and situational awareness capability, as well as collaborative decision-making and the emergence of collective intelligence. It also demonstrated multidimensional application value in empowering interdisciplinary thematic learning, offering effective tools to address these structural challenges. The realization of educational multi-agents empowering interdisciplinary thematic learning required the adoption of a sequential guidance-driven in design, a parallel role-based coordination empowerment in implementation, and a feedback-based adaptive intervention in evaluation. When applying educational multi-agents to empower interdisciplinary thematic learning, it was necessary to rely on the multi-agents collaboration architecture to maximize system efficiency, create a localized knowledge ecosystems and practical scenarios with multi-intelligent agents, design a human-machine collaboration model centered on learners, establish a closed loop of competency evaluation and intervention supported by multi-agents, and ensure the resource foundation and ethical norms for the operation of multi-agents.更多还原