Algorithmic monoculture in hiring: when one AI rejects the same candidates everywhere
What is algorithmic monoculture
Algorithmic monoculture occurs when many employers rely on the same AI vendor to screen job applicants. If a single vendor’s algorithm mediates hiring across an entire industry, a rejection at one company mechanically entails rejection at another, because the same model evaluates every application. Researchers at Stanford’s Human-Centered AI Institute described this as a new structural risk in the labor market: not just individual discrimination, but a system that shuts the same people out everywhere.
The evidence
A landmark study published in 2026 by the Stanford HAI team followed 3.4 million people submitting 4 million applications to 1,700 job postings across 150 employers and 11 industry sectors, all screened by a single third-party vendor. The findings were sobering. Applying the US Equal Employment Opportunity Commission’s four-fifths rule, which flags discrimination when one group is recommended at less than 80 percent of the rate of the most-recommended group, the researchers found that 26 percent of Black applicants and 15 percent of Asian applicants applied to positions where the AI discriminated against their racial group. If the AI had recommended Black and Asian candidates at the same rate as the most-favored group, 40,000 more applications would have advanced to the next stage.
Critically, these disparities were invisible in aggregate. When all applications were pooled together, the discrimination appeared to cancel out. Only when each position was analyzed separately did the bias emerge. The researchers recommend that regulators measure adverse impact per position, not averaged across all jobs.
Systemic rejection
The monoculture produces another effect: systemic rejection. Ten percent of applicants who submitted four applications were rejected from all positions. This rate is higher than what would be expected if each employer made decisions independently. For comparison, the largest prior study of hiring decisions, which sent 83,000 applications to 108 Fortune 500 firms, found that systemic rejection rates were accurately predicted by the baseline of independent decisions. The algorithmic hiring data systematically diverged from that baseline.
The mechanism is straightforward. When one vendor’s model screens applicants for multiple employers, a low score at one company predicts a low score at another. Even when different models are used, the researchers found that some applicants are systemically rejected across all positions they apply to. The rate of systemic rejection falls exponentially as applicants apply to more positions, but it decays more slowly than would be expected by chance.
Legal and regulatory response
The problem has attracted regulatory attention. The European Union’s AI Act classifies AI systems used in recruitment as high risk, requiring conformity assessments and ongoing monitoring. In the United States, New York City’s Local Law 144, effective July 2023, requires bias audits of automated employment decision tools. In June 2026, a federal judge in California ruled that Workday, whose AI-powered screening software is used by virtually all Fortune 500 companies, must face class-action claims alleging that its algorithms discriminated against Black applicants, women, and older workers. This case represents the first federal class action broadly targeting algorithmic decision-making in hiring.
What needs to change
The core problem is opacity. AI screening tools are pervasively adopted, highly consequential, and opaque to the public. Independent research into algorithmic hiring is essential, but it remains difficult because vendors do not share data. The Stanford researchers recommend three policy directions: measure adverse impact per position rather than in aggregate, monitor algorithmic monoculture as a structural risk, and create new pathways for independent research. Without transparency, evidence-based AI policy is impossible.
Oznaczenie treści AI
Ten articuł powstał z udziałem sztucznej inteligencji: research oraz tekst przygotował i zredagował agent AI. Dane, daty i cytaty przytoczone w tekstem pochodzą ze zweryfikowanych źródeł.
