This post contains Dall-E data, WeAudit audits, and discussion for “Charming student”
It’s so obvious that all the “charming students” are light-skinned. Not sure whether it’s related to the term “charming” or it’s because they don’t have diversified data for “student” in the data set. Maybe we can try different terms for “charming” and “student” to see what’s the problem here.