Gender and Racial Bias in Visual Question Answering Datasets

TL;DR

This paper investigates gender and racial biases in VQA datasets, revealing harmful stereotypes and underrepresented groups, and proposes solutions to mitigate these biases to improve fairness in vision-and-language models.

Contribution

The study provides a comprehensive analysis of biases in five VQA datasets and introduces strategies to reduce harmful stereotypes during dataset creation and use.

Findings

Answer distributions differ significantly between questions about women and men.

Gender-stereotypical samples are present and potentially harmful.

Race-related attributes are underrepresented, with discriminatory samples identified.

Abstract

Vision-and-language tasks have increasingly drawn more attention as a means to evaluate human-like reasoning in machine learning models. A popular task in the field is visual question answering (VQA), which aims to answer questions about images. However, VQA models have been shown to exploit language bias by learning the statistical correlations between questions and answers without looking into the image content: e.g., questions about the color of a banana are answered with yellow, even if the banana in the image is green. If societal bias (e.g., sexism, racism, ableism, etc.) is present in the training data, this problem may be causing VQA models to learn harmful stereotypes. For this reason, we investigate gender and racial bias in five VQA datasets. In our analysis, we find that the distribution of answers is highly different between questions about women and men, as well as the existence of detrimental gender-stereotypical samples. Likewise, we identify that specific race-related attributes are underrepresented, whereas potentially discriminatory samples appear in the analyzed datasets. Our findings suggest that there are dangers associated to using VQA datasets without considering and dealing with the potentially harmful stereotypes. We conclude the paper by proposing solutions to alleviate the problem before, during, and after the dataset collection process.