Gendered Information in Resumes and its Role in Algorithmic and Human Hiring Bias

Abstract

Listen

Translate article

We study the extent and nature of gendered information in resumes, and its role in hiring bias using a predictive modeling approach. We first train a series of machine learning (ML) models to inductively learn gendered information in resumes using a matched sample set of resumes from technology firms (348k resumes). We then use these models to quantify the extent and nature of gendered information in resumes and develop a measure of gender incongruence -- a predictive measure of how much the self-presented gender characteristics in the resume deviate from the gender of the candidate. Using this measure of gender incongruence along with historical hiring data from technology firms, we test whether applicants whose resume gender characteristics deviate from their actual gender (i.e. male resumes with feminine characteristics, female resumes with masculine characteristics) are less likely to receive a callback. Our results show that (1) there is a significant amount of gendered information in resumes -- even among applicants that apply to the same job opening with similar job-relevant characteristics, ML models can learn to differentiate between genders with a high degree of accuracy (AUROC=0.81). (2) This gendered information has an impact on hiring outcomes -- women whose resume characteristics deviate from the norms of their own gender are less likely to receive a callback after controlling for job-relevant characteristics. We discuss these findings in light of algorithmic and human bias in hiring.