可信学习分析构建:风险桎梏、理论纾解与实践解蔽  

Constructing Trustworthy Learning Analytics:Risky Issues,Theoretical Construction,and Practical Approach

作　　者：武法提[1,2] 李坦 夏志文 Wu Fati;Li Tan;Xia Zhiwen(Engineering Research Center of Digital Learning and Education Public Services,Ministry of Education,Beijing 100875;Faculty of Education,School of Educational Technology,Beijing Normal University,Beijing 100875)

机构地区：[1]北京师范大学教育学部教育技术学院,北京100875 [2]北京师范大学数字学习与教育公共服务教育部工程研究中心,北京100875

出　　处：《远程教育杂志》2024年第5期12-22,共11页Journal of Distance Education

基　　金：国家自然科学基金面上项目“同步直播课堂中基于多模态数据的学习者专注度评估及其演化机制研究”(项目编号:62177008)的研究成果。

摘　　要：可信学习分析技术的发展已成为全球研究者共同关注的焦点。国内外学者已就其内涵意蕴、价值逻辑等开展广泛探讨,但也存在关键问题不明、理论边界不清和实践路径模糊等问题。为此,研究以关键问题提炼和理论框架构建为先导,以实践路径的探索和实施为主线,对可信学习分析技术的教育应用进行分析。在实践问题方面,研究分析了阻碍可信学习分析实现的关键因素,包括数据漂移现象所导致的算法失准、样本分布失衡现象导致的算法偏见以及算法黑盒现象导致的决策结果失信等一系列问题。在理论研究方面,研究结合理论与实践问题分析,建立起面向学习环境、数据治理、建模分析和干预反馈四要素的可信学习分析理论框架。随后,研究依托该理论框架,结合相关研究方法和技术,构建了可信学习分析的实践路径,如结合价值敏感设计的伦理调适等。研究为可信学习分析研究从理论到实践提供系统化的解决思路。The development of trustworthy learning analytics technology has become a central focus for researchers globally.Both domestic and international scholars have explored its connotations and underlying value logic extensively.However,several challenges remain,including unclear core issues,ambiguous theoretical boundaries,and vague practical applications.This study aims to address these challenges by constructing a theoretical framework and exploring the practical potentials of trustworthy learning analytics in education.Practically,this study identifies the key barriers to implementing trustworthy learning analytics,such as algorithmic inaccuracies due to data drift,biases from imbalanced sample distribution,and the loss of trust caused by the“black box”nature of algorithms.Theoretically,the study combines analyses of theoretical and practical concerns to establish a comprehensive framework for trustworthy learning analytics,focusing on four key elements:learning environments,data governance,modeling,and intervention feedback.Building on this framework,the study proposes practical solutions,including ethical adaptation through value-sensitive design,to systematically bridge the gap between theory and practice in applying trustworthy learning analytics within educational research.

关 键 词：可信学习分析 可解释人工智能 算法公平 

分 类 号：G420[文化科学—课程与教学论]