Bias and Fairness in Algorithmic Hiring Systems

Algorithms are becoming increasingly prevalent in the hiring process. Whether it is a recruiter using LinkedIn's recommendation algorithm to find potential candidates or a hiring manager utilizing a resume screening algorithm to shortlist candidates, algorithms are increasingly used to assist human hiring decisions. These algorithms afford exciting opportunities for improving the efficiency of the hiring process but also pose several challenges along the lines of bias and fairness. This dissertation aims to investigate algorithmic hiring systems, with a particular emphasis on issues of bias, fairness, and diversity. The first chapter examines the interplay between algorithmic fairness constraints and human decision-making in hiring, highlighting the need for algorithms to be complementary to human decision-making. The second chapter studies how supply and demand-side choices in LinkedIn talent sourcing contribute to occupational segregation, contextualizing algorithmic hiring in the broader hiring process. The third chapter demonstrates how advances in machine learning algorithms can provide insight into the mechanisms underlying hiring bias. Finally, the fourth chapter builds on these findings and investigates the design and evaluation of fair resume screening algorithms.