在线协作知识建构情境下多智能体促进观点涌现的机制研究  

A Study of the Mechanisms Underlying Multi-Agent Facilitation of Idea Emergence in Online Collaborative Knowledge Building

作　　者：朱珂[1,2] 卞茗慧 张瑾[1] Zhu Ke;Bian Minghui;Zhang Jin(Faculty of Education,Henan Normal University,Xinxiang Henan 453007;Henan Province Collaborative Innovation Center for Intelligent Education,Henan Normal University,Xinxiang Henan 453007)

机构地区：[1]河南师范大学教育学部,河南新乡453002 [2]河南师范大学智能教育河南省协同创新中心,河南新乡453002

出　　处：《远程教育杂志》2025年第2期74-84,共11页Journal of Distance Education

基　　金：2024年河南省教育科学规划重大课题“河南推进教育数字化战略的模式与路径研究”(项目编号:2025JKZB06)的研究成果。

摘　　要：在线协作知识建构作为一种社会性学习形式,以观点涌现作为核心驱动,能够促进学习者之间的深度互动与知识建构。然而,观点涌现的质量与深度并非自然生成,而是受到学习者背景知识、团队协作频率、任务复杂性,以及技术支持环境等多重因素的影响。基于大语言模型技术的多智能体(LLM-MA)在协作过程中凭借其实时引导与反馈机制,展现出优化观点涌现过程的显著潜力。现有关于LLM-MA促进观点涌现的研究多聚焦于静态结果分析,较少关注观点生成的动态过程及其质量提升。因此,本研究以“社会知识建构互动分析模型”为理论基础,采用“共享信息、发现冲突、意义协商、知识构建、应用反思”五阶段协作流程,并结合过程挖掘分析与内容分析,系统探讨了基于大语言模型的多智能体在学习者在线协作知识建构中对观点涌现的作用效果。研究结果表明,基于大语言模型的多智能体不仅能够优化观点涌现的过程结构与效率,还能显著提升观点的深度与清晰度,而传统协作方式在观点涌现的广度上更具优势,表现出较强的探索性与创造力。最后,本研究针对教师介入、技术开发及学习者的社会性参与提出具体建议,旨在通过提供更加智能化的支持,优化在线协作知识建构中的观点涌现,推动大语言模型与教育实践的深度融合,全面提升学习者的协作效率与知识建构质量。Online collaborative knowledge building(OCKB),as a form of social learning,is fundamentally driven by the emergence of ideas,which facilitates deep interaction and meaningful knowledge construction among learners.However,the quality and depth of idea emergence do not occur spontaneously;rather,they are influenced by multiple factors,including learners’prior knowledge,collaboration intensity,task complexity,and the technology-enhanced learning environment.Multi-agent systems based on large language models(LLM-MA)have demonstrated significant potential in enhancing idea emergence through real-time guidance and feedback.Existing research on LLM-MA has primarily focused on static outcome analyses,with limited attention to the dynamic processes underlying idea generation and its quality improvement.Therefore,guided by the Interaction Analysis Model of Social Knowledge Construction,this study adopts a five-phase collaborative framework,which includes information sharing,conflict identification,meaning negotiation,knowledge construction,and reflective application.By integrating process mining and content analysis,the study systematically examines the effects of LLM-MA on idea emergence within OCKB contexts.The findings indicate that LLM-MA not only optimizes the structure and efficiency of idea emergence but also significantly enhances the depth and clarity of learners’ideas.In contrast,traditional collaborative approaches appear to foster greater breadth in idea emergence,characterized by stronger exploratory and creative potential.Based on these findings,the study proposes recommendations concerning teacher facilitation,technological development,and learners’social engagement.These suggestions aim to provide more intelligent support for enhancing idea emergence in OCKB,promote the meaningful integration of large language models into educational practice,and improve both collaborative effectiveness and the quality of knowledge construction.

关 键 词：在线协作知识建构 大语言模型 多智能体 观点涌现 

分 类 号：G420[文化科学—课程与教学论]