Avoiding Disparity Amplification under Different Worldviews

TL;DR

This paper compares different fairness definitions in machine learning under various assumptions about data bias, highlighting which ones prevent disparity amplification and proposing a new fairness notion aligned with a more realistic worldview.

Contribution

It provides a theoretical comparison of fairness criteria under different worldviews and introduces a new fairness concept suited for realistic data bias assumptions.

Findings

Demographic parity and equalized odds prevent disparity amplification under certain worldviews.

Predictive parity and calibration are insufficient to prevent disparity amplification.

A new fairness notion is proposed based on a more realistic worldview.

Abstract

We mathematically compare four competing definitions of group-level nondiscrimination: demographic parity, equalized odds, predictive parity, and calibration. Using the theoretical framework of Friedler et al., we study the properties of each definition under various worldviews, which are assumptions about how, if at all, the observed data is biased. We argue that different worldviews call for different definitions of fairness, and we specify the worldviews that, when combined with the desire to avoid a criterion for discrimination that we call disparity amplification, motivate demographic parity and equalized odds. We also argue that predictive parity and calibration are insufficient for avoiding disparity amplification because predictive parity allows an arbitrarily large inter-group disparity and calibration is not robust to post-processing. Finally, we define a worldview that is more realistic than the previously considered ones, and we introduce a new notion of fairness that corresponds to this worldview.