算法公平:教育人工智能算法偏见的逻辑与治理

Algorithmic Fairness:Logical Levels and Governance Focus of Algorithmic Bias in Educational Artificial Intelligence

算法公平被视作人工智能领域的核心伦理问题。教育人工智能同样面临算法偏见、算法歧视等伦理风险。通过系统性回顾2013一2023年教育人工智能算法偏见的57篇文献,本研究发现,从总体架构看,教育人工智能算法偏见研究主要分概念性理论研究、教育场景应用研究与算法检测设计研究三类;从以算法偏见为核心厘清算法伦理风险、算法歧视、教育公平三层概念的逻辑层次看,研究样本在算法偏见的种类、成因和治理原则与方法方面存在共性和明确的指向。本研究最后以种类和成因为基础,提出未来教育算法偏见研究的五个方向,以期促进教育的算法公平,优化人工智能教育应用环境,推动教育人工智能生态系统的健康发展。

The issue of algorithmic fairness is seen as an ethical issue at the heart of the field of artificial intelligence.Educational Al also faces ethical risks such as algorithmic bias and algorithmic discrimination.A systematic review of 57 papers on algorithmic bias in educational Al from 2013 to 2023 found the research on algorithmic bias in educational AI is mainly divided into three categories:Conceptual theoretical research,research on the application of educational scenarios,and research on the design of algorithmic detection.The review also found that from the logical hierarchy of clarifying the ethical risks of algorithmic ethics,algorithmic discrimination,and educational fairness were the core of algorithmic bias.In addition,the review discovered that from the logical hierarchy of the three conceptual layers,the research samples have commonalities and clear directions in terms of the types,causes,and governance principles and methods of algorithmic bias.Based on the review results,the article proposes five directions for the development of future research on algorithmic bias in the field of education:Slowing down the evolution from bias to discrimination,promoting algorithmic fairness in education,optimizing the environment of Al educational applications,and promoting the healthy development of the educational AI ecosystem.