Unequal Representations: Analyzing Intersectional Biases in Word Embeddings Using Representational Similarity Analysis

TL;DR

This paper introduces a novel method using representational similarity analysis to detect intersectional biases in word embeddings, revealing biases against Black women that reflect complex social discrimination patterns.

Contribution

It presents a new approach for analyzing intersectional biases in word embeddings, specifically highlighting biases against Black women using representational similarity analysis.

Findings

Black women are represented as less feminine than White women

Black women are represented as less Black than Black men

Embeddings reflect intersectional social biases

Abstract

We present a new approach for detecting human-like social biases in word embeddings using representational similarity analysis. Specifically, we probe contextualized and non-contextualized embeddings for evidence of intersectional biases against Black women. We show that these embeddings represent Black women as simultaneously less feminine than White women, and less Black than Black men. This finding aligns with intersectionality theory, which argues that multiple identity categories (such as race or sex) layer on top of each other in order to create unique modes of discrimination that are not shared by any individual category.