Algorithmic monocultures in hiring create racial disparities, Stanford study finds

A new Stanford study published at FAccT 2026 reveals how algorithmic monocultures—where multiple employers use screening algorithms from the same vendor—systematically disadvantage certain racial groups. Analyzing a dataset of 3 million applicants submitting 4 million applications, researchers Bommasani, Bana, Creel, Jurafsky, and Liang found clear racial disparities. Under U.S. employment discrimination standards, 14.74% of applications from Asian applicants and 25.87% from Black applicants were submitted to positions that adversely impact those groups. The homogeneity of outcomes is stark: 4% of applicants who applied to 10 positions were recommended for rejection from all, a rate significantly higher than random chance would predict.

To understand the mechanism, the researchers leveraged the deterministic replicability of hiring algorithms to simulate what would happen if applicants applied to all positions. They found that applicants would need to apply extremely broadly to increase their chances of having an application reviewed by a human. The study underscores a growing concern in AI ethics: when the same few algorithm vendors dominate hiring tools, they create a monoculture that amplifies bias and reduces opportunity for individual candidates. The authors suggest that employers should diversify their screening tools or mandate human review thresholds to mitigate these effects.