【作 者】：
杜修平， 王崟羽， 宋亚昕， 高玲泽
【关 键 词 】：
导学智能体； 大语言模型； 生成式意义建构； 高阶思维能力； PLS-SEM
【栏 目】：
其他
【中文摘要】：
生成式学习在实践中存在学习者参与主动性低、缺乏实时指导、个性化学习过程难以实施和生成式学习产物质量差等问题，大模型驱动的导学智能体凭借其多模态交互、动态认知诊断等特性，有望与生成式学习深度融合突破上述局限。为此，研究基于生成式意义建构框架构建导学智能体促进生成式意义建构理论模型，并结合眼动、EEG及问卷量表深入探究导学智能体对生成式意义建构过程及结果的影响和机制。研究发现，学习者在生成式意义建构过程中使用导学智能体能使大脑保持“低放松、高投入”状态，优化视觉认知行为，提高生成式学习产物质量；在学习结果上，学习者知识保持与迁移成绩显著提升，但导学智能体未能有效促进学习者的高阶思维能力，反而削弱问题解决能力。PLS-SEM分析发现，借助导学智能体提高生成式学习产物的质量是生成式意义建构成功的关键。最后，基于研究结果提出针对性建议，旨在破解生成式学习与AI融合的潜在教育风险，为生成式学习数智化发展提供理论支持和实践指导。
【英文摘要】：
Generative learning in practice has several challenges, including low learner engagement, lack of real-time guidance, difficulties in implementing personalized learning processes, and poor quality of generative learning products. Large model-driven tutoring agents, with their multimodal interaction and dynamic cognitive diagnosis, are expected to deeply integrate with generative learning and overcome these limitations. Therefore, based on the generative sense-making framework, this study constructs a theoretical model of tutoring agents facilitating generative sense-making, and combines eye-tracking, EEG and questionnaires to deeply explore the impact and mechanism of tutoring agents on the process and outcomes of generative sense-making. The study has found that when learners use tutoring agents in the process of generative sense-making, their brains can maintain a state of "low relaxation and high engagement", optimize visual cognitive behaviors, and improve the quality of generative learning outcomes. In terms of learning outcomes, learners demostrate significant improvements in knowledge retention and knowledge transfer. However, tutoring agents fail to effectively promote learners' higher-order thinking skills and even weaken their problem-solving abilities. PLS-SEM analysis further shows that improving the quality of GLA products with the assistance of tutoring agents is the key to successful generative sense-making. Finally, based on these findings, targeted suggestions are put forward to address the potential educational risks in the integration of generative learning and AI, providing theoretical support and practical guidance for the digital intelligence development of generative learning.