儿童教育中多智能体系统的应用:技术演进、作用机制与治理框架OA
Application of multi-agent system in children's education:technological evolution,mechanism of action,and governance framework
人工智能(artificial intelligence,AI)在教育领域的应用,已从单一答题工具逐步转向角色协同与流程协同模式.该文以学前与小学教育为研究范围,检索有关智能教学系统、教学代理、大语言模型智能体与多智能体系统(multi-agent system,MAS)的核心综述和代表性研究,采用结构化叙述综述方法,分析技术演进、应用场景、作用机制和治理边界.研究发现,MAS能够依托任务拆分、角色配合、同伴互动模拟及多模态过程评价,服务于阅读启蒙、数学探究、课堂协作、教师辅助及家校沟通等教育场景;当前相关研究仍以短期研究、准实验与原型验证为主.儿童教育领域的 MAS,应定位为遵循儿童发展规律、由教师主导并依托成人介入调控的人机协同学习基础载体.其有效应用需依托教育场景适配、数据最小化采集、人工在环核验及权责可追溯管理,实际推行可采用成人端优先布局、从低风险场景切入、分级逐步授权的推进方式.
The application of artificial intelligence(AI)in education has gradually evolved from isolated question-answering tools to modes of role collaboration and workflow orchestration.Focusing on preschool and primary education,this review synthesizes core reviews and representative studies on intelligent tutoring systems,pedagogical agents,large language model agents,and multi-agent system(MAS).Adopting a structured narrative review approach,it analyzes the technological evolution,application scenarios,mechanisms of action,and governance boundaries of these technologies.The results show that MAS can support educational scenarios,including reading enlightenment,mathematical inquiry,classroom collaboration,teacher assistance,and home-school communication through task decomposition,role cooperation,peer interaction simulation,and multimodal process assessment.Current relevant studies still focus primarily on short-term researches,quasi-experiments,and prototype validations.In the field of children's education,MAS should be positioned as a fundamental carrier for human-AI collaborative learning that conforms to children's developmental laws,is teacher-led,and is regulated by adult mediation.Its effective application relies on educational scenario adaptation,minimal data collection,human-in-the-loop verification,and traceable management of rights and responsibilities.For practical promotion,priority deployment on the adult side,initiation in low-risk scenarios,and gradual hierarchical authorization are recommended.
郑博利;王慧;戴永亮;金志娟;胡孟晗
华东师范大学医学与健康研究院(中国 上海 200241)万达信息股份有限公司(中国 上海 201112)万达信息股份有限公司(中国 上海 201112)上海交通大学医学院附属上海儿童医学中心(中国 上海 200127)华东师范大学信息与电子工程学院(中国 上海 200241)
多智能体系统儿童教育教育智能体大语言模型人机协同
multi-agent systemchildren's educationeducational agentlarge language modelhuman-AI collaboration
《教育生物学杂志》 2026 (3)
163-170,175,9
评论