Lookahead Counterfactual Fairness

TL;DR

This paper introduces lookahead counterfactual fairness (LCF), a new fairness framework that considers the downstream effects of ML predictions on individuals' future outcomes, addressing limitations of traditional counterfactual fairness.

Contribution

It proposes the concept of LCF, provides theoretical conditions for its satisfaction, and develops an algorithm to implement this fairness notion, extending it to path-dependent scenarios.

Findings

The proposed method effectively enforces LCF on synthetic data.

Experiments show improved fairness in real-world datasets.

Theoretical analysis clarifies conditions for LCF satisfaction.

Abstract

As machine learning (ML) algorithms are used in applications that involve humans, concerns have arisen that these algorithms may be biased against certain social groups. \textit{Counterfactual fairness} (CF) is a fairness notion proposed in Kusner et al. (2017) that measures the unfairness of ML predictions; it requires that the prediction perceived by an individual in the real world has the same marginal distribution as it would be in a counterfactual world, in which the individual belongs to a different group. Although CF ensures fair ML predictions, it fails to consider the downstream effects of ML predictions on individuals. Since humans are strategic and often adapt their behaviors in response to the ML system, predictions that satisfy CF may not lead to a fair future outcome for the individuals. In this paper, we introduce \textit{lookahead counterfactual fairness} (LCF), a fairness notion accounting for the downstream effects of ML models which requires the individual \textit{future status} to be counterfactually fair. We theoretically identify conditions under which LCF can be satisfied and propose an algorithm based on the theorems. We also extend the concept to path-dependent fairness. Experiments on both synthetic and real data validate the proposed method.