A Friendly Face: Do Text-to-Image Systems Rely on Stereotypes when the Input is Under-Specified?
This addresses bias and diversity issues in AI-generated images for the general public and researchers, but it is incremental as it builds on existing stereotype literature.
The study investigated whether text-to-image systems produce demographic biases in response to under-specified prompts with social attributes, finding that images often reflect stereotypes similar to those in literature, though inconsistently across models.
As text-to-image systems continue to grow in popularity with the general public, questions have arisen about bias and diversity in the generated images. Here, we investigate properties of images generated in response to prompts which are visually under-specified, but contain salient social attributes (e.g., 'a portrait of a threatening person' versus 'a portrait of a friendly person'). Grounding our work in social cognition theory, we find that in many cases, images contain similar demographic biases to those reported in the stereotype literature. However, trends are inconsistent across different models and further investigation is warranted.