Examining Gender and Racial Bias in Large Vision-Language Models Using a Novel Dataset of Parallel Images
8 February 2024
Computers and Society
Kathleen C. Fraser,
New dataset of AI-generated 'parallel' images varying only in gender and race
Images cover everyday scenarios like occupations and activities
Queried vision-language chatbots exhibit gender and racial bias
Bias manifests in occupation stereotypes and crime associations
Examining biases in vision-language chatbots
The paper presents a new dataset of AI-generated images that are identical except for the gender and race of the person depicted. By querying vision-language chatbots with these images, the authors uncover differences in how the chatbots describe people based on perceived gender and race.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper