学业成就与学业风险的预测——基于学习分析领域中预测指标的文献综述

Prediction of academic performance and risk: a review of literature on predicative indicators in learning analytics

学习分析作为一个从数据中建构意义的研究领域,在过去几年的发展中备受学界关注。学习分析领域的核心问题之一是如何利用数据预测学习者的学业成功或者失败?围绕这一问题,国内外学者开展了大量实证研究,取得了丰富的研究成果。但是,预测指标研究的相关综述却存在一定局限性,如忽视指标适用的学习场所和情境、模糊指标匹配的学习任务类型和参与主体,或是有些综述缺失了领域内的代表性学者、研究和应用。因此,本文通过系统的文献检索和综述,从预测指标适用的学习场所和任务类型出发,梳理了倾向性指标、人机交互指标和人际交互指标三种类型的常用预测指标。本文详细地介绍了过往学业表现、初始知识、学习驱动力、正面或负面学习行为、学习者情感状态、知识表征事件、人际交互频次、社群意识等一系列得到广泛验证的关键预测指标,并将按照"学校场所和工作场所"和"个体学习和群体学习"两个维度划分的四个象限,在每个象限中选取一个典型的学习分析系统进行剖析,这些典型系统是Signals系统、SNAPP系统、Learn-B系统和Cohere系统。本文最后总结了预测分析相关研究的特点和趋势,并指明了未来研究与实践的注意事项和潜在的研究方向。

As a research area to construct meaning from data, learning analytics has drawn great attention from academics in its development. One of the key issues of learning analytics, researched both domestically and abroad in comprehensive empirical studies and with profound research result, is how to predict learners' learning success or failure. However, there have been limitations in the literature review of studies of predicative indicators neglecting the applicable situations or contexts, blurring the task types and suitable participants and leaving out representative researchers, their studies and practice. Through systematic literature retrieval and review, focusing on learning contexts and task types, this study analyzes three types of commonly used predictive indicators, namely, dispositional indicators, human-machine interaction indicators, and human-human interaction indicators. It gives a detailed account of crucial predictive indicators proven to be effective, namely, past academic performance, initial knowledge, learning motivation, positive or negative learning behaviors, learner's emotional status, knowledge representation events, human-human interaction frequency, sense of community, etc. This study also analyzes one typical learning analysis system in each of the four quadrants developed based on the two dimensions of"at school or in the workplace"and"individual learning or group learning". Finally, future development and research trends are proposed.