Word-Level Explanations for Analyzing Bias in Text-to-Image Models

TL;DR

This paper presents a method to identify which words in prompts influence bias in text-to-image models, helping to understand and mitigate societal stereotypes in generated images.

Contribution

It introduces a novel word influence scoring method based on masked language models to analyze bias in text-to-image generation.

Findings

Effectively identifies bias-inducing words in prompts

Reveals societal stereotypes in generated images

Applicable to models like Stable Diffusion

Abstract

Text-to-image models take a sentence (i.e., prompt) and generate images associated with this input prompt. These models have created award wining-art, videos, and even synthetic datasets. However, text-to-image (T2I) models can generate images that underrepresent minorities based on race and sex. This paper investigates which word in the input prompt is responsible for bias in generated images. We introduce a method for computing scores for each word in the prompt; these scores represent its influence on biases in the model's output. Our method follows the principle of \emph{explaining by removing}, leveraging masked language models to calculate the influence scores. We perform experiments on Stable Diffusion to demonstrate that our method identifies the replication of societal stereotypes in generated images.