Algorithmic hiring tools risk hiding and amplifying workplace inequalities: while promised as efficient and objective, recruitment AIs can reproduce bias through opaque models and institutional legitimation, requiring interdisciplinary scrutiny and shared regulatory responsibility.

Main Finding

AI adoption in hiring is reshaping organizational inequality by increasing the invisibility and perceived legitimacy of selection processes. Across disciplines, the authors find a dominance of technical framings that treat AI as a technosolution to human bias, which narrows attention to individual-level fixes while obscuring structural drivers of inequality. They introduce the concept of "algorithmically‑mediated inequality regimes" to capture how AI can reproduce, conceal, and reconfigure entrenched hiring inequalities through black‑box decisioning, expanded data extraction, and growing institutional trust in algorithmic tools.

Key Points

• Multidisciplinary landscape: Research on AI in hiring spans computing/data science (CS), HR/management/organizations (HRMOS), social sciences (SS), and law (LS). Each discipline emphasizes different concerns and methods, producing important but often siloed debates.
• Asymmetries in framing: Technical fields often focus on bias detection and algorithmic fairness (debiasing), while social-science and legal literatures emphasize structural, organizational, and power dynamics. The rising dominance of technical perspectives risks marginalizing structural analyses.
• Algorithmic invisibility and legitimacy: AI intensifies opacity ("black box") in decision-making and benefits from a cultural and organizational legitimacy that makes algorithmic choices harder to contest.
• Hiring as a gatekeeper: Because hiring determines who enters organizations, algorithmic changes to hiring can create cascading cumulative effects on careers, income, and social mobility.
• Surveillance capitalism and data access: Widespread data collection and extraction in hiring (e.g., video, voice, social traces) amplify employer power and raise privacy concerns.
• Regulatory lag and contestation: Law and policy responses are fragmented, uneven across jurisdictions, and lag behind technological deployment, leading to unclear accountability.
• Conceptual contribution: Extends Acker’s (2006) “inequality regimes” to propose "algorithmically‑mediated inequality regimes" that foreground how AI both conceals and reproduces organizational inequalities.
• Prescriptive directions: Calls for interdisciplinary "chain of knowledge" (integrating technical, social, and legal expertise) and multi‑stakeholder "chain of responsibility" (developers, employers, regulators) to govern AI hiring systems.

Data & Methods

• Hybrid review design combining:
  • Scoping review (Arksey & O’Malley style) to map the breadth of literature at the intersection of AI, hiring, and inequality.
  • Problematizing review (Alvesson & Sandberg) to question underlying assumptions, surface blind spots, and generate new conceptual framings.
• Search and screening:
  • Initial search across multiple databases (Google Scholar, ProQuest, Sociological Abstracts, SSCI, Sage, SSRN) using keywords around AI/algorithms, hiring/HR, and (in)equality/fairness/bias produced 389 items.
  • Blind screening via Rayyan by a four‑author team classified items; 212 publications were retained as "highly relevant."
• Deep review and selection:
  • A "supercorpus" of 97 readings was constructed (57 from scoping + 40 additional targeted items) emphasizing conceptually generative works.
  • Team reflexivity: each reviewer acted as a primary reader for one disciplinary cluster and a secondary reader rotating across others to build cross‑disciplinary bridges.
• Coding/mapping:
  • Publications grouped into four disciplinary clusters: CS, HRMOS, SS, LS; coded for questions, concepts, assumptions.
• Limitations of method:
  • Scoping component does not perform formal quality appraisal typical of systematic reviews.
  • Selection in the problematizing phase is interpretive and shaped by the team’s positionalities; emphasis is on conceptual depth rather than representativeness or meta‑analytic synthesis.

Implications for AI Economics

• Labor market allocation and matching
  • Algorithmic hiring changes the information environment and matching frictions: more automated screening can improve some efficiency metrics but may systematically exclude groups, altering equilibrium match outcomes and potentially lowering aggregate human capital utilization.
  • Hidden selection mechanisms create unobserved heterogeneity in access to job queues, complicating estimation of returns to skills and signaling models.
• Distributional and dynamic inequality effects
  • Because hiring is an entry point with cumulative career effects, algorithmic bias can magnify income and opportunity inequality over cohorts and across generations, with broader implications for social mobility and aggregate inequality.
  • AI‑driven exclusion may produce negative externalities (e.g., skill depreciation among excluded groups), altering labor supply responses and long‑run productivity.
• Information asymmetries and market power
  • Increased opacity in selection algorithms raises information asymmetries between firms and workers, and among firms and regulators. Large platforms/vendors that supply hiring algorithms may accrue market power, influencing hiring standards and norms.
• Welfare and regulatory economics
  • The paper highlights a need for policy tools (transparency mandates, auditability, data governance, anti‑discrimination enforcement adapted to algorithms) and economic analysis of their costs and benefits, including compliance costs, impacts on hiring efficiency, and redistribution effects.
  • Fragmented regulation increases uncertainty for firms and may generate uneven competitive effects across jurisdictions.
• Research and measurement priorities for economists
  • Develop micro‑founded models incorporating algorithmic screening as a mechanism that conditions access to jobs and career trajectories.
  • Empirically estimate causal effects of algorithmic hiring on outcomes (employment rates, wages, turnover, promotion) using field experiments, audit studies, and quasi‑experimental designs.
  • Quantify distributional impacts and long‑run welfare consequences, including general equilibrium feedbacks.
  • Cost‑benefit analyses of regulatory interventions (e.g., transparency/audit rules, data minimization requirements, vendor liability) that account for enforcement frictions.
• Design and governance implications
  • Encourage interdisciplinary evaluation teams (economists, computer scientists, organizational scholars, lawyers) to assess hiring tools’ labor‑market impacts before wide deployment.
  • Promote standards for algorithmic audits and data provenance that make economic evaluation and accountability feasible.
• Policy caution: Efficiency gains touted by vendors may come with hidden distributional harms; economic policy should balance productivity aims with equity and information‑provision mandates.

Limitations and caveats - The article is a conceptual and multidisciplinary literature review rather than new empirical estimation; implications for AI economics require empirical validation. - The review’s interpretive selection (problematizing phase) emphasizes conceptual insights over exhaustive representativeness; economists should combine these insights with quantitative evidence.