学习科学视角下的隐马尔可夫模型——联结理论与数据的学习过程分析方法  

Hidden Markov Models from a learning sciences perspective:analysing learning processes by linking theory and data

作　　者：张鹏 苏晗宇 张媛媛 尚俊杰[1] ZHANG Peng;SU Hanyu;ZHANG Yuanyuan;SHANG Junjie(Graduate School of Education,Peking University,Beijing 100871,China;Development Planning Division,Education Department,Jiangxi,Nanchang 330038,China)

机构地区：[1]北京大学教育学院,北京100871 [2]江西省教育厅发展规划处,江西南昌330038

出　　处：《教学研究》2025年第1期1-10,共10页Research in Teaching

基　　金：国家自然科学基金2023年面上项目(62377001)。

摘　　要：学习过程分析是学习分析的核心议题,但常因理论依据与分析技术之间的融合不佳,而限制了对教育意义的深入解释。隐马尔可夫模型(HMM)可以有效结合学习科学理论经验和模型构建,为解决此困境提供了一种简单、透明、可解释的方案。因此,从HMM分析方法切入,详细介绍了HMM的基本组成及其在教育中的重要应用价值,说明如何基于HMM建模学习过程和预测学习成果,强调了学习科学理论与数据相结合的重要性。通过多个案例分析,展示了HMM在游戏化学习、学业成就预测和自我调节学习过程分析中的典型应用。最后,总结了HMM在教育研究中分析组别差异、多阶段过程和多通道过程的三大方向与策略,为未来更有教育意义的学习分析研究提供理论与实践指导。Analyzing the learning process is a core focus in learning analytics,but the lack of effective integration between theoretical foundations and analytical techniques often limits the depth of educational insights.The Hidden Markov Model(HMM)offers a powerful,transparent,and interpretable approach to bridge this gap by integrating learning science theory with model construction.This article introduces HMM from an analytical perspective,detailing its basic components and highlighting its important applications in education.It explains how HMM can model learning processes and predict learning outcomes,emphasizing the value of combining theoretical frameworks from Learning Sciences with data-driven methods.Through several case studies,the article illustrates HMM’s typical applications in areas such as game-based learning,academic achievement prediction,and self-regulated learning process analysis.Finally,the article outlines three key directions for using HMM in educational research:analyzing group differences,multi-stage processes,and multi-channel processes.These strategies provide theoretical and practical guidance for advancing learning analytics research with greater educational impact.

关 键 词：学习过程 学习分析 隐马尔可夫模型 学习科学 研究方法 

分 类 号：G43[文化科学—教育学]