Word-Level Explanations for Analyzing Bias in Text-to-Image Models
This addresses bias issues in AI-generated content, which is crucial for fairness in applications like art and datasets, though it is incremental as it builds on existing explanation techniques.
The paper tackles the problem of bias in text-to-image models by developing a method to identify which words in input prompts cause underrepresentation of minorities in generated images, demonstrating its effectiveness on Stable Diffusion to reveal societal stereotypes.
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.