ChatGPT is biased against resumes with credentials that imply a disability—but it can improve
Which disabilities did the modified ChatGPT still show bias against?
The modified ChatGPT still showed bias against cerebral palsy, as it was the only disability among the six tested that did not rank higher than resumes that didn't mention disability. The other five implied disabilities - deafness, blindness, autism, the general term "disability", and depression - improved after the customization with written instructions directing the tool not to be ableist.
How did UW researchers modify ChatGPT to reduce ableism?
UW researchers modified ChatGPT by customizing the tool with written instructions directing it not to be ableist. They used the GPTs Editor tool, which allowed them to customize GPT-4 with these instructions without requiring any code. This approach aimed to make the AI model work with disability justice and DEI principles, reducing ableist biases in its outputs.
What methods did researchers use to imply disabilities in resumes?
The researchers created six enhanced CVs, each implying a different disability by including four disability-related credentials: a scholarship, an award, a diversity, equity and inclusion (DEI) panel seat, and membership in a student organization. They used one of the study's authors' publicly available curriculum vitae (CV) as the base.