Algorithmic Monocultures in Hiring
A study finds shared hiring algorithms can amplify adverse impact and systemic rejection across employers.
This study analyzes 3.4 million real applicants and 4 million applications across 156 U.S. employers. It finds position-level racial adverse impact that aggregate analysis can obscure, especially affecting Black and Asian applicants. The authors also show that reliance on a single vendor can create homogeneous outcomes and systemic rejections, calling for stronger audits, surveillance, and researcher access.
"Algorithmic Monocultures in Hiring" is an empirical study about the risks of hiring AI, with authors including Rishi Bommasani, Sarah H. Bana, Kathleen A. Creel, Dan Jurafsky, and Percy Liang. The background of the study is that a large number of U.S. employers already use algorithms to screen job applicants, and many companies rely on a small number of the same vendors. When multiple employers use the same algorithm—or the same vendor's hiring algorithm—to evaluate candidates, the authors call this an "algorithmic monoculture": on the surface, different companies are each hiring independently, but in reality job seekers may be repeatedly judged by the same type of model logic.
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on Hacker News (AI keywords) →Summaries are AI-generated; the original article is authoritative.