A Stanford-led study has found that artificial intelligence (AI) hiring tools used by major employers are disproportionately screening out Black and Asian job applicants, raising fresh concerns about algorithmic bias in recruitment systems, reported The Financial Times.
The study, which analysed 4 million job applications across 156 employers using the Pymetrics hiring platform, found evidence of what researchers described as “systemic rejection” linked to AI-driven assessments.
Researchers said candidates often had to apply for at least 25 jobs to be almost certain of progressing beyond the initial screening stage.
The research is considered the largest examination so far of AI hiring algorithms and their impact on recruitment outcomes.According to the findings, around one in 10 job roles in the dataset showed “adverse impact” against Black applicants, while one in 20 roles negatively affected Asian candidates. The term “adverse impact” is used by US federal agencies when selection rates for certain racial or ethnic groups fall significantly below others.
The study also found that several companies were using identical algorithmic screening models, meaning candidates rejected by one employer could automatically fail at others using the same system.
Researchers identified 42 shared models across different employers. About 4 per cent of applicants who applied for 10 different roles were rejected by the platform’s algorithm every time — a higher rate than would occur randomly.
Pymetrics, now owned by Harver, uses online game-based assessments to evaluate traits such as risk-taking, reaction speed, trust and social behaviour. Candidates whose profiles closely match existing top-performing employees are recommended to move forward in the hiring process.
Kathleen Creel, co-author of the study and assistant professor at Northeastern University, said the dominance of a single AI vendor across industries could amplify flaws across the entire hiring ecosystem.
The researchers noted that while the findings raise concerns, the results may not apply to all AI hiring systems, as different platforms use different screening methods, including résumé analysis and video interviews.
The study adds to growing global scrutiny of automated hiring tools as companies increasingly rely on AI to manage large volumes of job applications.