【作 者】：
张 琪， 王红梅
【关 键 词 】：
学习投入； 多模态； 数据表征； 关键技术； 学习分析
【栏 目】：
理论探讨
【中文摘要】：
多模态数据建模已成为洞察学习规律的新范式。研究梳理了学习投入的概念演进与评测方法，从经典教育学理论、教育神经科学、具身认知理论以及量化学习视角阐释了学习投入的内在机制、研究范式、研究方法与技术前景，构建多模态数据表征学习投入的理论基础。在此基础上，分析了学习投入的发生机制，提出从情感状态、认知参与状态以及与学习环境互动产生的行为综合表征学习投入的观点。建立包含学习者瞬时行为数据、内容交互数据、情境互动数据的分析框架，围绕学习行为建模、模态传感器建模、算法模型以及新技术的介入四个方面讨论多模态数据建模的关键技术。通过多模态数据的整合分析，结合机器学习方法，可分析学习投入的细粒度指标以及在不同场景中的建模过程，超越独立数据源难以整合关联的问题，最终实现探索教育智能时代的学习规律、改善学习的目的。
【英文摘要】：
Multiple-modality data modeling has become a new paradigm for working out learning rules. This paper sorts out the conceptual evolution and evaluation methods of learning engagement, explains the internal mechanism, research paradigm, research method and technical prospect of learning engagement from the perspective of classical pedagogical theory, educational neuroscience, embodied cognition theory and quantitative learning, and then constructs a theoretical basis of learning engagement of multiple-modality data representation. On this basis, this paper analyzes the mechanism of learning engagement, and puts forward that learning engagement is a comprehensive characterization of emotional state, cognitive participation state and behaviors generated by interacting with learning environment. Then, an analysis framework consisting of learners' instantaneous behavior data, learner-content interaction data, learner-context interaction data is established and the key technologies of multiple-modality data modeling from four aspects of learning behavior modeling, modal sensor modeling, algorithm modeling and new technology intervention is discussed. Through the analysis of multiple-modality data, combined with machine learning method, both the fine-grained indicators of learning engagement and the modeling process in different scenarios can be analyzed. Finally, the problem which independent data sources are difficult to integrate associations can be solved and the learning rules can be explored in the era of educational intelligence.