基于教育智能体自学的县中学习机制构建与干预研究——以高一数学为例

Construction and Intervention Study of Learning Mechanisms for County High School Students Based on Educational Agent Self-Study—A Case Study of Grade 10 Mathematics

摘要: 随着县域普通高中自习课对高质量学习支持与过程治理需求的增加，教育智能体在课堂中的应用受到关注。然而，县域普通高中场景普遍存在低质量交互、过程不可见与策略性应付等问题，限制了智能体赋能自学的实际成效。针对上述需求，本研究提出一种“教育专用智能体 + 学习过程治理”的县域普通高中自学支持方案，构建“目标设定–交互学习–反向输出评价”的闭环学习机制，并配套分阶段的反作弊与质量纠偏规则，以实现大班条件下较为可执行、可审计的课堂治理。研究采用准实验对比与过程数据追踪相结合的方法，在高一两个平行班的有效样本(n = 86)中开展为期6周干预，以数学一模与二模成绩评估成效，并结合系统对话日志、教师干预记录与学生反向输出文本分析交互行为与阶段演进特征。结果显示时间 × 组别交互显著(F = 6.145, p = 0.015, η² = 0.068)，同时提炼出“新接触–适应–策略性规避–引导修正–良好习惯”的五阶段演进模型与关键干预点。研究为县域普通高中人工智能辅助自学提供了具有一定可操作性的实施思路与初步实证参考。

Abstract: With the growing demand for high-quality learning support and process governance in self-study periods at county-level ordinary high schools, the application of educational agents in classrooms has attracted increasing attention. However, such settings commonly suffer from low-quality interactions, invisible learning processes, and strategic compliance behaviors, which limit the practical effectiveness of agent-supported self-study. To address these challenges, this study proposes a self-study support scheme that combines education-specific agents with learning-process governance, constructing a closed-loop mechanism of “goal setting - interactive learning - reverse-output evaluation”, together with phased anti-cheating and quality-correction rules to support relatively executable and auditable classroom governance under large-class conditions. Using a quasi-experimental comparison combined with process-data tracking, a 6-week intervention was conducted with an effective sample of 86 students from two parallel Grade 10 classes. Mathematics Mock Exam I and II scores were used to assess outcomes, while system dialogue logs, teacher intervention records, and students’ reverse-output texts were analyzed to examine interaction behaviors and stage evolution. Results showed a significant Time × Group interaction (F = 6.145, p = 0.015, η² = 0.068). A five-stage evolution model of “initial contact-adaptation-strategic avoidance-guided revision-good habits” and corresponding intervention points were identified. The findings provide a practically relevant implementation approach and preliminary empirical evidence for AI-assisted self-study in county-level ordinary high schools.