摘要
算法歧视关涉三个具体的伦理问题:算法公平、算法身份的污名化与隐私。其原因在于:一是数据中预先存在的偏见所导致的算法歧视;二是使用算法本身可能是一种歧视;三是算法决策中数据的抽样偏差及其所设置权重的不同也有可能导致算法歧视。基于此,至少可从技术进路、哲学进路与法治进路三个方向探究规避算法歧视的可能性,进而实现算法公平。
Algorithmic discrimination involves three ethical issues:algorithmic fairness,the stigmatization of algorithmic identity,and privacy.The reasons are as follows:First,algorithmic discrimination can be caused by the preexisting bias in the data;Second,using algorithms itself may be a form of discrimination;Third,the sampling bias of the data and different weight settings in the algorithmic decision-making may also lead to algorithmic discrimination.Therefore,algorithmic discrimination can be avoided from three directions:the technical approach,the philosophic approach and the approach of rule of law,so as to realize algorithmic fairness.
作者
刘培
池忠军
LIU Pei;CHI Zhongjun(School of Marxism,China University of Mining and Technology,Xuzhou,Jiangsu,221116)
出处
《自然辩证法通讯》
CSSCI
北大核心
2019年第10期16-23,共8页
Journal of Dialectics of Nature
关键词
算法歧视
算法公平
算法身份污名化
Algorithmic Discrimination
Algorithmic Fairness
Stigmatization of algorithmic identity
