【摘要】 随着ChatGPT、DeepSeek等大模型的快速发展，生成式人工智能技术已深度融入教育。教育生态正从传统数字教育形态跃迁为人机共生、交往理性的数智教育生态，成为“师—机—生”三元互动和物理、文化、数字三元交融的复合场域。本研究聚焦数智教育生态四个核心维度的范式变革：教学主体从单一走向多元，教师角色由知识传授者变为学习设计者，学生逐渐成为主动探索者，智能体作为教育“准主体”深度参与教学；知识观从静态走向动态，教学组织从单一走向混合；学习方式实现认知过程外显化，强调批判性使用与创造性应用知识；教学评价由结果导向转向多维整合，由静态测量转向动态适应。基于教学案例，本研究借助DeepSeek双模型架构设计了五阶段教学流程，开发了四类功能性教育智能体，并提出差异化智能体应用策略，以期为数智教育生态下人机协同教学提供实践范式与理论支持。更多还原

【Abstract】 Since the emergence of ChatGPT, generative artificial intelligence technology has deeply integrated into various domains of education. Education transforms traditional digital education into a digital-intelligent education ecosystem with human-machine symbiosis and rational communication. This transformation needs to establish a “ teacher-machine-student” three-way interactive structure and a complex physical, cultural, and digital integration field. This study focuses on four core dimensional paradigm shifts in the digital-intelligent education ecosystem. First, in the new ecosystem, the teaching subjects evolve from singular to plural, changing teachers’ roles from knowledge transmitters to learning designers. The students gradually become active explorers and intelligent agents participating deeply in teaching as educational “ quasi-subjects.” The view of knowledge also changes from static to dynamic and teaching organization changes from single to mixed with learning methods realizing the externalization of cognitive processes, emphasizing critical use and creative application. Finally, educational evaluation shifts from result-oriented to multidimensional integration and from static measurement to dynamic adaptation. Based on teaching cases, this study utilizes DeepSeek’s dual-model architecture to design a five-stage teaching process, develops four types of functional educational intelligent agents, and proposes differentiated intelligent agent application strategies, providing practical paradigms and theoretical support for human-machine collaborative teaching in the digital-intelligent education ecosystem.更多还原