Inscribing Diversity Policies in Algorithmic Hiring Systems: Theory and Empirics

Abstract

We study the downstream effects of diversity policies inscribed as algorithmic fairness constraints in Human+AI hiring systems. We present, solve, and empirically estimate a 2-stage hiring model consisting of: (1) a screening algorithm, which selects and shortlists candidates from a pool of applicants, and (2) an unbiased hiring manager, who hires from the shortlist. We consider the equal selection fairness constraint (i.e., a diversity policy), which constrains the screening algorithm to shortlist an equal number of men and women. We solve this model analytically and show that under certain conditions, even when both the algorithm and the hiring manager are unbiased, the fairness constraint can be ineffective in increasing the diversity of the hires. The more correlated the screening algorithm’s and the hiring manager’s assessment criteria are, the less effective the equal selection constraint becomes in increasing workforce diversity. Based on our theoretical findings, we propose a screening algorithm that would increase the effectiveness of the equal selection constraint. We empirically test our theoretical predictions using hiring data from eight IT firms and show via counterfactual policy simulation that the equal selection constraint in the shortlist can only improve the gender diversity of hires by a modest amount and not up to parity. We benchmark these results against our proposed algorithmic design and other commonly used fairness constraints and show the effectiveness of the proposed design in increasing workforce diversity. 

Keywords: algorithmic hiring, human-ai interaction, algorithmic fairness, diversity