Authors: Hari Nagakoteswar Tripurari
Abstract: AI-based hiring systems promise substantial efficiency gains in candidate screening and selection, yet their rapid diffusion has occurred alongside mounting regulatory scrutiny regarding algorithmic bias, decision transparency, and employer legal accountability. This paper examines how organizations redesign algorithmic screening systems in response to emerging AI governance frameworks and labor-market fairness expectations, including the EU AI Act, New York City Local Law 144, and evolving U.S. Equal Employment Opportunity Commission guidance. Employing a quasi-experimental difference-in-differences design across 164 organizational hiring units spanning multiple jurisdictions and five years (2021–2026), the study evaluates changes in applicant diversity, hire quality, employee turnover, candidate trust, and legal risk exposure associated with algorithmic redesign. The study further compares five hiring system architectures — opaque black-box AI, explainable AI (XAI) with limited oversight, human-in-the-loop (HITL) review of opaque models, integrated XAI-HITL systems, and fully manual baselines — across all outcome dimensions. Regression results indicate that explainability and human oversight intensity each independently and interactively predict improved applicant diversity (β = 0.21 and 0.16, respectively, both p < .001), reduced legal risk exposure (β = –1.38 and –1.62, both p < .001), and enhanced institutional legitimacy, with integrated XAI-HITL architectures achieving superior outcomes across all five performance and fairness dimensions relative to opaque systems and to fully manual baselines. Thematic analysis of 39 executive interviews identifies six organizational themes, including a recurring 'human-in-the-loop paradox' in which oversight mechanisms erode through automation complacency absent deliberate design safeguards. The paper develops the Algorithmic Hiring Legitimacy Framework, integrating efficiency, fairness, explainability, and institutional legitimacy as interdependent design dimensions, and contributes novel empirical evidence to HR analytics, responsible AI, and information systems governance research.
DOI: http://doi.org/10.5281/zenodo.20840456
