Towards Equal Opportunity Fairness through Adversarial Learning

TL;DR

This paper introduces an augmented adversarial training method that explicitly models equal opportunity fairness in NLP, leading to improved bias mitigation and better performance-fairness trade-offs.

Contribution

It proposes a novel augmented discriminator for adversarial training that explicitly incorporates equal opportunity considerations in bias mitigation.

Findings

Significant improvement over standard adversarial debiasing methods.

Enhanced performance-fairness trade-off in experiments.

Effective modeling of equal opportunity in NLP bias mitigation.

Abstract

Adversarial training is a common approach for bias mitigation in natural language processing. Although most work on debiasing is motivated by equal opportunity, it is not explicitly captured in standard adversarial training. In this paper, we propose an augmented discriminator for adversarial training, which takes the target class as input to create richer features and more explicitly model equal opportunity. Experimental results over two datasets show that our method substantially improves over standard adversarial debiasing methods, in terms of the performance--fairness trade-off.