Approaches to Improve Fairness when Deploying AI-based Algorithms in Hiring - Using a Systematic Literature Review to Guide Future Research

Abstract

Algorithmic fairness in Information Systems (IS) is a concept that aims to mitigate systematic discrimination and bias in automated decision making. However, previous research argued that different fairness criteria are often incompatible. In hiring, AI is used to assess and rank applicants according to their fit for vacant positions. However, various types of bias also exist for AI-based algorithms (e.g., using biased historical data). To reduce AI's bias and thereby unfair treatment, we conducted a systematic literature review to identify suitable strategies for the context of hiring. We identified nine fundamental articles in this context and extracted four types of approaches to address unfairness in AI, namely pre-process, in-process, post-process, and feature selection. Based on our findings, we (a) derived a research agenda for future studies and (b) proposed strategies for practitioners who design and develop AIs for hiring purposes.