【作 者】：
魏艳涛， 徐 琦， 师亚飞， 刘清堂
【关 键 词 】：
经验之塔； 多智能体； 生成式人工智能； 智能导学系统； 系统构建
【栏 目】：
其他
【中文摘要】：
基于生成式人工智能的导学系统可通过模拟专业导师增强学习效果，但是现有系统重直接经验、轻间接经验，重教学辅导、轻学习支持，对学生认知发展支持不足。为解决上述挑战，研究融合生成式人工智能技术与经验之塔理论，设计开发了生成式多智能体导学系统，通过直接经验与间接经验相统一的学习模式和“教—学—导”三元智能体来支持学生的学习。研究发现：系统拥有高质量的教学对话生成能力，兼具知识高效掌握和技术高接受度的双重优势；面向直接经验学习模式的“指导智能体”综合表现最好，面向间接经验学习模式的“交互型教师”角色最有助于学生获得学术成功。研究证实了直接经验与间接经验的系统化融合能够优化学生的知识掌握，三元智能体架构可满足多元学习需求，以期对智能导学系统的开发和应用提供有效参考。
【英文摘要】：
Generative Artificial Intelligence (GAI)-based Tutoring Systems can enhance learning outcomes by simulating professional mentors. However, existing systems often prioritize direct experience over indirect experience, focus more on instructional guidance than learning support, and inadequately support students' cognitive development. To address these challenges, this study develops a Generative Multi-agent Tutoring System by integrating GAI with Cone of Experience theory. This system facilitates students' learning by creating a learning model that unifies direct experience and indirect experience, and a "teaching-learning-guiding" tri-agent. The findings indicate that the system has the ability to generate high-quality instructional dialogues and has the dual advantages of efficient knowledge acquisition and high technology acceptance. The guiding agent for direct experience learning presents optimal overall performance, and the interactive teacher for indirect experience learning contributes to students' academic success most. The study confirms that systematic integration of direct and indirect experiences optimizes students' knowledge acquisition, and the tri-agent architecture successfully accommodates diverse learning needs. These results provide effective references for the development and application of intelligent tutoring system.