Assessing Intersectional Bias in Representations of Pre-Trained Image Recognition Models

TL;DR

This paper investigates biases in pre-trained image recognition models, focusing on facial images and intersectional variables like age, race, and gender, revealing age-related differentiation and ethnicity and gender associations.

Contribution

It introduces a method to analyze intersectional biases in ImageNet classifiers using probes and visualization, highlighting specific biases in facial recognition.

Findings

Models differentiate ages more strongly.

Ethnicity and gender biases are present in middle-aged groups.

Biases vary across intersectional categories.

Abstract

Deep Learning models have achieved remarkable success. Training them is often accelerated by building on top of pre-trained models which poses the risk of perpetuating encoded biases. Here, we investigate biases in the representations of commonly used ImageNet classifiers for facial images while considering intersections of sensitive variables age, race and gender. To assess the biases, we use linear classifier probes and visualize activations as topographic maps. We find that representations in ImageNet classifiers particularly allow differentiation between ages. Less strongly pronounced, the models appear to associate certain ethnicities and distinguish genders in middle-aged groups.