Data Management for Causal Algorithmic Fairness

TL;DR

This paper emphasizes the importance of causal reasoning in data management to address fairness issues in machine learning, highlighting the need for causal approaches over associational ones.

Contribution

It distinguishes between associational and causal fairness, advocating for causal reasoning and reviewing data management techniques applicable to causal fairness.

Findings

Causal fairness requires understanding data generation processes.

Existing data management techniques can be adapted for causal fairness.

Future research opportunities in causal data management are identified.

Abstract

Fairness is increasingly recognized as a critical component of machine learning systems. However, it is the underlying data on which these systems are trained that often reflects discrimination, suggesting a data management problem. In this paper, we first make a distinction between associational and causal definitions of fairness in the literature and argue that the concept of fairness requires causal reasoning. We then review existing works and identify future opportunities for applying data management techniques to causal algorithmic fairness.