Exploring Gender Bias Beyond Occupational Titles

TL;DR

This paper investigates gender bias beyond occupational stereotypes by introducing a new dataset and framework that measure and interpret contextual gender biases, demonstrating their presence across diverse datasets including Japanese.

Contribution

The authors present GenderLexicon, a novel dataset, and a framework for estimating and explaining contextual gender bias in language models.

Findings

Gender biases exist beyond occupational stereotypes.

The framework effectively estimates and explains gender bias.

Bias is confirmed across multiple datasets, including Japanese.

Abstract

In this work, we investigate the correlation between gender and contextual biases, focusing on elements such as action verbs, object nouns, and particularly on occupations. We introduce a novel dataset, GenderLexicon, and a framework that can estimate contextual bias and its related gender bias. Our model can interpret the bias with a score and thus improve the explainability of gender bias. Also, our findings confirm the existence of gender biases beyond occupational stereotypes. To validate our approach and demonstrate its effectiveness, we conduct evaluations on five diverse datasets, including a Japanese dataset.