The Authority of "Fair" in Machine Learning

TL;DR

This paper advocates for a normative, community-inclusive approach to defining fairness in machine learning, reviewing current literature and encouraging broader debate on ethical standards.

Contribution

It introduces a normative framework for fairness in ML and discusses how to incorporate diverse community perspectives into fairness standards.

Findings

Highlights the importance of normative fairness definitions

Reviews current Fair ML literature and its implications

Suggests community engagement for ethical consensus

Abstract

In this paper, we argue for the adoption of a normative definition of fairness within the machine learning community. After characterizing this definition, we review the current literature of Fair ML in light of its implications. We end by suggesting ways to incorporate a broader community and generate further debate around how to decide what is fair in ML.