Automating inequality? Gender bias in AI-mediated labor markets
In the authors' words
Generative AI is rapidly entering labor markets, with large language models (LLMs) now used to help write CVs and evaluate applicants. We study the risks this creates in a series of audit experiments in which applicant gender is randomly assigned while all other information is held constant. We find that LLMs generate gender-stereotypical CVs and exhibit gender bias in screening scores and wage recommendations. Screening scores track occupational gender segregation, penalizing women in male-dominated occupations and favoring them in female-dominated ones. Wage recommendations track occupational gender wage gaps when gender is explicit in the prompt, though this gradient weakens under name-based gender signaling and is sensitive to the LLMs being used. These disparities arise through two channels: evaluation bias, when identical CV content is evaluated differently by gender, and generation bias, when gendered content produced by the model affects later evaluation. Because the same systems shape both CV content and its subsequent assessment, these channels can reinforce one another. Taken together, our findings identify measurable risks of algorithmic amplification in AI-mediated job search.
In plain words AI summary
Researchers tested whether AI language models show gender bias in job applications. They randomly assigned gender to identical applicants in experiments. The models produced CVs with gender stereotypes and gave biased scores. Women were penalized in male-dominated jobs and favored in female-dominated ones. Bias came from how CVs were written and how they were judged, which can reinforce each other.
Written by Kimi K2.5 using only the abstract above. Jev checked that every sentence is supported by the abstract sentences it cites, and removed any that were not. Point at a sentence to highlight its source.
Appeared: Wednesday, November 18. Proceedings of the National Academy of Sciences. Peer-reviewed journal.