Algorithmic Hiring and the Limits of Procedural Fairness

Abstract

Automated screening tools are now widely used to shortlist job applicants, and their vendors often present procedural transparency as a guarantee of fairness. This article examines that claim. We analyse 46 hiring systems used by mid-sized employers, combining audits of published decision rules with interviews of recruiters and applicants. We find that making a procedure visible does not prevent unequal outcomes when the data used to train or tune the tool reflect earlier patterns of exclusion. Applicants also read transparency differently from designers: many treated detailed scoring explanations as evidence of rigidity rather than fairness, and few felt able to contest a result. We propose a framework that separates procedural, distributive and interactional fairness in algorithmic selection, and we show where each can fail. The paper concludes with practical recommendations on auditing, human review and meaningful routes of appeal, and identifies limits that no procedure alone can remove.