Balanced Datasets Are Not Enough: Estimating and Mitigating Gender Bias in Deep Image Representations

TL;DR

This paper introduces a framework to measure and reduce gender bias in deep image recognition models, revealing that balanced datasets alone do not prevent bias amplification during training.

Contribution

It demonstrates that models amplify gender associations beyond dataset biases and proposes an adversarial method to mitigate this bias effectively.

Findings

Models amplify gender associations beyond dataset biases.

Balanced datasets do not prevent bias amplification.

Adversarial training reduces gender bias while maintaining accuracy.

Abstract

In this work, we present a framework to measure and mitigate intrinsic biases with respect to protected variables --such as gender-- in visual recognition tasks. We show that trained models significantly amplify the association of target labels with gender beyond what one would expect from biased datasets. Surprisingly, we show that even when datasets are balanced such that each label co-occurs equally with each gender, learned models amplify the association between labels and gender, as much as if data had not been balanced! To mitigate this, we adopt an adversarial approach to remove unwanted features corresponding to protected variables from intermediate representations in a deep neural network -- and provide a detailed analysis of its effectiveness. Experiments on two datasets: the COCO dataset (objects), and the imSitu dataset (actions), show reductions in gender bias amplification while maintaining most of the accuracy of the original models.