Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Human, Algorithm, or Both? Gender Bias in Human-Augmented Recruiting

Mesut Kaya, Toine Bogers

TL;DR

This work provides one of the first empirical comparisons of fairness across human, AI, and hybrid recruiting processes, offering evidence to inform the development of more equitable hiring practices and highlighting the importance of human oversight for mitigating bias in algorithmic hiring.

Abstract

Recent years have seen rapid growth in the market for HR technology and AI-driven HR solutions in particular. This popularity has also resulted in increased attention to the negative aspects of using AI to support hiring practices, such as the risk of reinforcing existing biases against vulnerable groups based on gender or other sensitive attributes. Combining human experience with AI efficiency in making recruiting and selection decisions has the potential to help mitigate these biases, but despite a considerable amount of research on fairness in algorithmic hiring, actual empirical evaluations comparing the fairness of human, AI, and human-augmented decision-making remain scarce. In this study, we address this gap by presenting a quantitative analysis of gender bias across three scenarios of a real-world recruitment platform: (1) recruiters searching a CV database manually for relevant candidates, (2) AI-driven matching between candidates and jobs, and (3) a combination of human and AI-driven recruiting. We find that human recruiters produce lists of candidates that are fairer in terms of gender than the AI-only solution, with more deliberation by humans resulting in fairer outcomes. However, the combination of human and AI-driven is more than the sum of its parts and produces the fairest candidate lists: interacting with the slate of recommended candidates first before manually searching for additional candidates has a beneficial effect on the gender fairness of the set of candidates that are viewed, clicked, and contacted afterwards. Our work provides one of the first empirical comparisons of fairness across human, AI, and hybrid recruiting processes, offering evidence to inform the development of more equitable hiring practices and highlighting the importance of human oversight for mitigating bias in algorithmic hiring.

Human, Algorithm, or Both? Gender Bias in Human-Augmented Recruiting

TL;DR

This work provides one of the first empirical comparisons of fairness across human, AI, and hybrid recruiting processes, offering evidence to inform the development of more equitable hiring practices and highlighting the importance of human oversight for mitigating bias in algorithmic hiring.

Abstract

Recent years have seen rapid growth in the market for HR technology and AI-driven HR solutions in particular. This popularity has also resulted in increased attention to the negative aspects of using AI to support hiring practices, such as the risk of reinforcing existing biases against vulnerable groups based on gender or other sensitive attributes. Combining human experience with AI efficiency in making recruiting and selection decisions has the potential to help mitigate these biases, but despite a considerable amount of research on fairness in algorithmic hiring, actual empirical evaluations comparing the fairness of human, AI, and human-augmented decision-making remain scarce. In this study, we address this gap by presenting a quantitative analysis of gender bias across three scenarios of a real-world recruitment platform: (1) recruiters searching a CV database manually for relevant candidates, (2) AI-driven matching between candidates and jobs, and (3) a combination of human and AI-driven recruiting. We find that human recruiters produce lists of candidates that are fairer in terms of gender than the AI-only solution, with more deliberation by humans resulting in fairer outcomes. However, the combination of human and AI-driven is more than the sum of its parts and produces the fairest candidate lists: interacting with the slate of recommended candidates first before manually searching for additional candidates has a beneficial effect on the gender fairness of the set of candidates that are viewed, clicked, and contacted afterwards. Our work provides one of the first empirical comparisons of fairness across human, AI, and hybrid recruiting processes, offering evidence to inform the development of more equitable hiring practices and highlighting the importance of human oversight for mitigating bias in algorithmic hiring.
Paper Structure (25 sections, 1 equation, 2 figures, 3 tables)

This paper contains 25 sections, 1 equation, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Human Oversight
  4. Fairness in Human Hiring
  5. Fairness in Algorithmic Hiring
  6. Methodology
  7. Experimental design
  8. Data
  9. Fairness Metrics
  10. Notation
  11. Demographic Parity
  12. Conditional Demographic (Dis)Parity
  13. Inferring gender
  14. Measuring fairness
  15. Human recruiting
  16. ...and 10 more sections

Figures (2)

  • Figure 1: A visualization of the recruiter workflow in the different conditions. Recruiters are first presented with the slate of recommended candidates ( ). They can choose to skip these recommendations, after which their slate of selected candidates will solely be based on their own searches ( ). If recruiters decide to engage with the slate of recommended candidates, their workflow is a combination of AI-based recommendation and manual searching ( ), which we further divide into the AI oversight phase ( ) and the manual searching phase after having interacted with the recommendations ( ).
  • Figure 2: These five plots visualize the changes in CDP ratios over time for the different scenarios: (a) , (b) , (c) , (d) , and (e) . Scenarios with human involvement show three CDP time series for the Viewed, Clicked, and Contacted candidates, while scenarios that involve AI show two CDP time series for the Recommended and Recommended Top-K candidates. The areas shaded in green in each plot correspond to the 80% heuristic originating from the EEOC's guidelines, showing a band of 20% on either side of the perfect CDP score of 1.0, representing perfect fairness. Red-shaded areas indicate CDP ratios outside of these bands.