摘要
算法偏见是社会偏见在算法信息社会的延伸,映射信息不平等。本文采用系统综述方法,对国内外算法偏见类型的实证研究进行梳理综合,在界定算法偏见的内涵和外延基础上,整合算法偏见的二元主体概念框架,并归纳算法偏见的三条形成链,包括人类智能系统内的偏见链、人工智能系统内的偏见循环圈、人类智能系统到人工智能系统的偏见作用路径。然后从功能主义、冲突论、还原论和马克思主义角度反思算法偏见研究的哲理基础。最后从信息科学视角构建技术、信息、用户之间的三元交互影响模型,讨论算法偏见在技术、信息、用户三要素互动过程中的存在形式,并发现算法偏见导致的三类不平等现象——信息呈现不平等、信息分布不均衡和新型数字不平等。研究结果深刻揭示了算法时代的信息社会问题,并赋予信息领域的传统不平等话语新的内涵与外延,为信息职业人员应对算法时代新挑战提供思路和参考。图3。表2。参考文献92。
More and more algorithms are being used to replace or assist humans in distributing important social goods and making automated decision-making,which has contributed greatly to social development.But algorithms are not entirely fair,and public concern about algorithmic bias has become a social issue as algorithmic bias is discovered and reported.This paper adopts a systematic review method to sort out and synthesize empirical research on algorithmic bias,aiming to answer three questions:What is algorithmic bias?What are the philosophy foundations of algorithmic bias?And what are the impacts of algorithmic bias in the perspective of information science?Therefore,on the basis of defining the connotation and extension of algorithmic bias,the dual subject conceptual framework of algorithmic bias—human intelligence system and artificial intelligence system—is summarized.It reveals the bias chain within the human intelligence system,the bias loop within the artificial intelligence system,and the bias action path from the human intelligence system to the artificial intelligence system.Algorithmic bias is essentially a philosophical topic,and reflection on its philosophical foundations can help people better understand the formation mechanism of algorithmic bias.The philosophical perspectives adopted by algorithmic bias reflection mainly include Functionalism(including Structural Functionalism and Machine Functionalism),Conflict Theories,Reductionism,and Marxism.Different perspectives have different interpretations of the forming mechanism of algorithmic bias.At last,a three-dimensional interaction model of technology,information,and users is constructed to discuss the impacts of algorithmic bias from the perspective of information science.Three major impacts of algorithm bias on information inequality and digital inequality are described:inequality of information presentation,unbalanced distribution of information,and a new form of digital inequality known as the algorithms divide.By deeply revealing the algorithmic social problems in the era of artificial intelligence,this research endows the traditional discourse of inequality in the information field with new connotation and extension,proposing possible references for information professionals to face new challenges.3 figs.2 tabs.92 refs.
作者
贾诗威
闫慧
JIA Shiwei;YAN Hui
出处
《中国图书馆学报》
CSSCI
北大核心
2022年第6期57-76,共20页
Journal of Library Science in China
基金
国家社会科学基金项目“数字中国背景下数字贫困消减行动研究”(编号:18BTQ007)的研究成果之一。
关键词
算法偏见
信息不平等
数字不平等
信息呈现
信息分布
Algorithmic bias
Information inequality
Digital inequality
Information presentation
Information distribution
