Publication on Multimodal Learning Analysis

Liu, Q., Li, X., Xie, K., Chang, B., & Zheng, X. (2022). Developments and Prospects of Empirical Research on Multimodal Learning Analysis. e-Education Research, 1, 71-78. DOI: 10.13811/j.cnki.eer.2022.01.009 (In Chinese)

多模态学习分析实证研究的发展与展望

Abstract
Multimodal learning analysis(MMLA)is a key technology for intelligently exploring the mechanism of effective learning. This study conducts a systematic review of 37 foreign empirical literature on task scenario design and the four processes of MMLA. It is found that the generative fields of multimodal datasets are mainly focused on developing cognition and less on cultivating affective values.Learning label annotation is mainly guided by computational science, but lacks theoretical guidance on behavioral correlation at different time scales. The prediction results pay more attention to the performance of learning behaviors than to the explanation of the process of mental development. Feedback from multimodal data analysis focuses on personalized learning support and neglects decision support. Future empirical studies should pay attention to effective learning and affective experience, integrate computational science and cognitive band theory, provide feedback support in collaboration with humanmachine advantage, conduct in-depth dialogues between MMLA system developers and stakeholders,continuously iterate the design and optimize the analysis system and application models to effectively promote the development of “AI+Education”.

摘要: 多模态学习分析(MMLA)是智能化探究有效学习发生机理的关键技术。研究对国外37篇实证文献的任务情境设计和MMLA的四个过程进行系统综述,梳理出多模态数据集的生成场域多以发展认知为主,少关注情感价值的培育;学习标签注释以计算科学指导为主,缺乏不同时间尺度行为关联的理论指导;预测结果多关注学习行为表现,轻心智发展的过程解释;多模态数据分析反馈聚焦个性化学习支持,忽视决策支持。未来实证研究发展应聚焦有效学习与情感体验,融合计算科学和认知带理论,协同人机优势提供反馈支持,开展MMLA系统开发者和利益相关者的深度对话,不断迭代设计与优化分析系统和应用模式,有效促进”人工智能+教育”的发展。

Key words
Multimodal Learning Analysis; Learning Behavior; Learning Label; Formal Learning Scenarios; Empirical Study;

关键词: 多模态学习分析; 学习行为; 学习标签; 正式学习情境; 实证研究;