Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Algorithmic Hiring and Diversity: Reducing Human-Algorithm Similarity for Better Outcomes

Prasanna Parasurama, Panos Ipeirotis

TL;DR

The paper addresses how algorithmic fairness constraints, specifically equal selection in shortlists, affect final hire diversity in a two-stage hiring process. It develops a theoretical framework where the screening score $Q^S$ and hiring manager score $Q^H$ are jointly modeled with true candidate quality $Q$, introducing a correlation parameter $\theta$ that governs diversity outcomes and hire quality; empirical validation uses nearly $8\times 10^5$ applicants from eight tech firms, estimating $\theta$ and gender differences, and conducting counterfactual simulations. The main finding is that higher $\theta$ (closer alignment between screener and manager) reduces the effectiveness of equal-shortlist diversity and can reduce hire quality, motivating a complementary screening algorithm that deliberately differs from managerial preferences yet remains predictive, which empirically improves gender diversity in final hires with limited quality loss. The work offers practical guidance for designing hiring pipelines: maximize screening accuracy while minimizing correlation with managerial assessments, or train screening to complement human judgments (e.g., adversarial or multi-objective approaches) to better achieve both diversity and quality in hires.

Abstract

Algorithmic tools are increasingly used in hiring to improve fairness and diversity, often by enforcing constraints such as gender-balanced candidate shortlists. However, we show theoretically and empirically that enforcing equal representation at the shortlist stage does not necessarily translate into more diverse final hires, even when there is no gender bias in the hiring stage. We identify a crucial factor influencing this outcome: the correlation between the algorithm's screening criteria and the human hiring manager's evaluation criteria -- higher correlation leads to lower diversity in final hires. Using a large-scale empirical analysis of nearly 800,000 job applications across multiple technology firms, we find that enforcing equal shortlists yields limited improvements in hire diversity when the algorithmic screening closely mirrors the hiring manager's preferences. We propose a complementary algorithmic approach designed explicitly to diversify shortlists by selecting candidates likely to be overlooked by managers, yet still competitive according to their evaluation criteria. Empirical simulations show that this approach significantly enhances gender diversity in final hires without substantially compromising hire quality. These findings highlight the importance of algorithmic design choices in achieving organizational diversity goals and provide actionable guidance for practitioners implementing fairness-oriented hiring algorithms.

Algorithmic Hiring and Diversity: Reducing Human-Algorithm Similarity for Better Outcomes

TL;DR

The paper addresses how algorithmic fairness constraints, specifically equal selection in shortlists, affect final hire diversity in a two-stage hiring process. It develops a theoretical framework where the screening score and hiring manager score are jointly modeled with true candidate quality , introducing a correlation parameter that governs diversity outcomes and hire quality; empirical validation uses nearly applicants from eight tech firms, estimating and gender differences, and conducting counterfactual simulations. The main finding is that higher (closer alignment between screener and manager) reduces the effectiveness of equal-shortlist diversity and can reduce hire quality, motivating a complementary screening algorithm that deliberately differs from managerial preferences yet remains predictive, which empirically improves gender diversity in final hires with limited quality loss. The work offers practical guidance for designing hiring pipelines: maximize screening accuracy while minimizing correlation with managerial assessments, or train screening to complement human judgments (e.g., adversarial or multi-objective approaches) to better achieve both diversity and quality in hires.

Abstract

Algorithmic tools are increasingly used in hiring to improve fairness and diversity, often by enforcing constraints such as gender-balanced candidate shortlists. However, we show theoretically and empirically that enforcing equal representation at the shortlist stage does not necessarily translate into more diverse final hires, even when there is no gender bias in the hiring stage. We identify a crucial factor influencing this outcome: the correlation between the algorithm's screening criteria and the human hiring manager's evaluation criteria -- higher correlation leads to lower diversity in final hires. Using a large-scale empirical analysis of nearly 800,000 job applications across multiple technology firms, we find that enforcing equal shortlists yields limited improvements in hire diversity when the algorithmic screening closely mirrors the hiring manager's preferences. We propose a complementary algorithmic approach designed explicitly to diversify shortlists by selecting candidates likely to be overlooked by managers, yet still competitive according to their evaluation criteria. Empirical simulations show that this approach significantly enhances gender diversity in final hires without substantially compromising hire quality. These findings highlight the importance of algorithmic design choices in achieving organizational diversity goals and provide actionable guidance for practitioners implementing fairness-oriented hiring algorithms.

Paper Structure

This paper contains 44 sections, 7 theorems, 12 equations, 12 figures, 12 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Theoretical Framework and Implications
  4. Model Setup
  5. Theoretical results
  6. Effects on hire diversity
  7. Effects on hire quality
  8. Implications for Algorithm Design
  9. Algorithmic Selection with Independent Parameters
  10. Balancing Predictive Accuracy and Managerial Complementarity
  11. Directly Minimizing Gender Differences in QĤ Scores
  12. Data and Empirical Methodology
  13. Data Description
  14. Empirical Strategy
  15. Step 1: Estimating Screening and Hiring Manager Scores.
  16. ...and 29 more sections

Key Result

Proposition 1

The effectiveness of the equal selection constraint ($p_h$) decreases as the correlation ($\theta$) between algorithmic scores and hiring manager scores increases.

Figures (12)

  • Figure 1: The correlation structure between $Q, Q^S, Q^H$
  • Figure 2: Female proportion of hires ($p_h$) vs. correlation parameter ($\theta$)
  • Figure 3: Female proportion of hires ($p_h$) vs. gender difference in correlation parameter ($\delta$)
  • Figure 4: Expected quality of hire ($E[Q_h]$) vs. $\theta$, $\theta^S$
  • Figure 5: Target variable options for training a screening algorithm $\mathcal{A}$.
  • ...and 7 more figures

Theorems & Definitions (10)

  • Proposition 1
  • Corollary
  • Proposition 2
  • Proposition 3
  • Proposition
  • proof
  • Proposition
  • proof
  • Proposition
  • proof