Research work · 2023
Race and Intersecting Inequalities in Machine Learning
Sharla Alegria
The Oxford Handbook of the Sociology of Machine Learning
Abstract
Abstract The central problem of race and gender inequality in machine learning (ML) reflects how inequalities of the past can be captured and reproduced in the content, solutions, and applications that use large amounts of data to make decisions. This chapter provides examples to explain four key areas of concern about race and gender inequalities in ML technology: (a) in the data itself, (b) in the ways data are used in algorithms, (c) in the ways algorithms are optimized, and (d) in the ways ML applications are used in the social world. The chapter further extends the discussion to examine issues of privacy and data collection and the inequalities that emerge around the human work needed to create and sustain highly data-intensive ML tools. Finally, the chapter explores strategies for improving fairness and equity.