Recruiters Follow AI's Biased Hiring Recommendations 90% of the Time, Research Says

"The researchers gave every resume the last name Williams-chosen because it's the third most common surname in the U.S. and almost perfectly split between Black and White Americans (47.68% vs. 45.75% according to Census data). Then they swapped in first names from a validated database. Nothing else on the resume changed. Experience, degrees, job history: all identical."

"The intersectional finding is the starkest single data point in either paper: When resumes with Black male names were compared directly against resumes with White male names, White male names were preferred in 100% of 27 bias tests. Zero exceptions across all three models and all nine occupations."

"Across all of them, the pattern held-the models treated White and male as the default, with other identities treated as deviations from it. (Interestingly, the bias wasn't based on occupational patterns. HR, for example, skews 76.5% female in the real-world workforce, but the models still favored male names for HR roles.)"

University of Washington researchers conducted experiments testing whether AI hiring tools exhibit racial and gender bias and whether human reviewers detect it. Three top-performing language models were tested on over 554 real resumes across 571 job descriptions spanning nine occupations. Researchers manipulated only candidate names while keeping all other resume details identical, using the surname Williams paired with first names from validated databases to test racial and gender bias. Across all nine occupations tested, models consistently treated White and male identities as default, with other identities treated as deviations. Most strikingly, White male names were preferred over Black male names in 100% of 27 bias tests across all three models and occupations, with zero exceptions. The bias persisted regardless of occupational gender patterns in the real workforce.

Read at TechRepublic