Counterfactual Fairness Is Basically Demographic Parity

Making fair decisions is crucial to ethically implementing machine learning algorithms in social settings. In this work, we consider the celebrated definition of counterfactual fairness. We begin by showing that an algorithm which satisfies fairness also demographic parity, a far simpler constraint. Similarly, show all satisfying parity can be trivially modified satisfy Together, our results indicate basically equivalent has important implications for growing body work on then validate theoretical findings empirically, analyzing three existing against simple benchmarks. find two benchmark outperform algorithms---in terms fairness, accuracy, and efficiency---on several data sets. Our analysis leads us formalize concrete goal: preserve order individuals within protected groups. believe transparency around ordering groups makes more trustworthy. By design, goal while do not.