A Moral Framework for Understanding of Fair ML through Economic Models of Equality of Opportunity

TL;DR

This paper unifies various algorithmic fairness notions under economic models of Equality of Opportunity, clarifies their moral assumptions, and introduces new fairness measures inspired by luck egalitarianism, with empirical implications.

Contribution

It provides a formal mapping of fairness notions to EOP, unifies them under a moral framework, and proposes new fairness measures based on luck egalitarian principles.

Findings

Many fairness definitions are special cases of EOP.

Moral assumptions underlying fairness notions are explicitly identified.

Using unfairness measures can harm disadvantaged groups.

Abstract

We map the recently proposed notions of algorithmic fairness to economic models of Equality of opportunity (EOP)---an extensively studied ideal of fairness in political philosophy. We formally show that through our conceptual mapping, many existing definition of algorithmic fairness, such as predictive value parity and equality of odds, can be interpreted as special cases of EOP. In this respect, our work serves as a unifying moral framework for understanding existing notions of algorithmic fairness. Most importantly, this framework allows us to explicitly spell out the moral assumptions underlying each notion of fairness, and interpret recent fairness impossibility results in a new light. Last but not least and inspired by luck egalitarian models of EOP, we propose a new family of measures for algorithmic fairness. We illustrate our proposal empirically and show that employing a measure of algorithmic (un)fairness when its underlying moral assumptions are not satisfied, can have devastating consequences for the disadvantaged group's welfare.