Reproduce “Algorithmic Monocultures in Hiring” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Rishi Bommasani, Sarah H. Bana, Kathleen A. Creel, Dan Jurafsky, Percy Liang
An empirical study investigated algorithmic monocultures in hiring, revealing that reliance on a single vendor's algorithms leads to racial adverse impact at the per-position level for Black and Asian applicants, and an increased rate of systemic rejection for individuals across multiple applications. Counterfactual simulations demonstrate that applicants must apply to significantly more positions to secure a single recommendation compared to a scenario with independent hiring decisions.
Key insights
- Aggregate fairness metrics reported by algorithmic vendors can obscure substantial racial adverse impact when assessed on a per-position basis, consistent with U.S. employment discrimination guidelines.
- Algorithmic monoculture drives outcome homogenization, resulting in higher rates of systemic rejection for applicants compared to what would be expected if hiring decisions were independent.
- Every applicant possesses traits that align with at least some algorithmic models, suggesting systemic rejection is a function of the algorithmic ecosystem rather than universal unsuitability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.