Fair AI in hiring: Experimental evidence on how biased hiring algorithms and different debiasing methods affect the quality and diversity of applicants

Employers increasingly use artificial intelligence (AI) to help screen job applicants. But the way algorithms are designed can make them highly susceptible to systematic biases against women in male-dominated fields, including the women who are best qualified for the jobs. Various methods for debiasing hiring algorithms are available to help ensure that the best-qualified men and women are all given fair consideration. However, little is known about which approach is most likely to appeal to applicants. To address this question, the author conducted a mock job application experiment that examined how decisions to apply for a competitive job differ depending on the debiasing approach used. The study showed that awareness of gender bias in an algorithm significantly deterred women from applying, including those who were most qualified for the job. It also demonstrated that all the debiasing approaches studied significantly increased the number of women applying without compromising the quality of the final applicant pool. An approach designed to give men and women an equal chance of being selected attracted the most female applicants and should be considered by employers who primarily wish to increase diversity in hiring. However, men and women in this study also perceived an algorithm that ignored gender as the fairest. Such gender-blind algorithms should be considered by employers who want to both improve numbers of female applicants and reinforce workers’ perceptions that hiring practices are fair.