"AI, Automation, and Labor Market Inequality: Algorithmic Hiring as a Second-Order Displacement Mechanism" by CeCe Grimberg

AI, Automation, and Labor Market Inequality: Algorithmic Hiring as a Second-Order Displacement Mechanism

CeCe Grimberg, Loyola University Maryland

Abstract: Accelerating developments in artificial intelligence and automation have magnified systemic anxieties surrounding workforce displacement. While technological innovation has historically created economic opportunities, access to employment is increasingly controlled by algorithmic hiring systems designed to optimize recruitment efficiency. This paper examines the intersection of these trends, arguing that automated screening technologies act as a second-order displacement mechanism that intensifies labor market inequality. Workers displaced by automation frequently develop non-linear career trajectories, characterized by employment gaps, cross-sector transitions, and alternative credentials. However, predictive hiring algorithms and applicant tracking systems are typically trained on historical hiring data that reward continuous, traditional career pathways while penalizing deviations. Consequently, displaced workers face systematic exclusion from reemployment opportunities, transforming temporary technological disruption into persistent economic marginalization. This dual exclusion mechanism disproportionately impacts vulnerable populations, leading to severe macroeconomic inefficiencies, reduced labor mobility, and exacerbated geographic and intergenerational inequality. Although algorithmic systems offer standardization and reduced processing costs, their opacity and scale magnify structural biases. To mitigate these effects, this paper proposes institutional and policy reforms, including mandated disparate-impact audits, the retention of human review for non-traditional applicants, and the regulation of algorithmic hiring as high-impact labor-market infrastructure. Ultimately, sustaining innovation-driven growth requires evolving labor market institutions to ensure equitable access to employment in an automated economy.

---

Introduction:

Rapid advances in artificial intelligence have intensified concerns surrounding automation-driven labor displacement and the future of employment. While technological innovation has historically generated productivity gains alongside new forms of economic opportunity, the contemporary labor market differs in one critical respect: access to employment is increasingly mediated by algorithmic hiring systems. Artificial intelligence is no longer confined to automating manufacturing processes or repetitive administrative tasks; it now plays a central role in determining which workers gain access to interviews, advancement opportunities, and reentry into the labor market itself. As firms adopt AI-driven applicant tracking systems (ATS), resume-ranking software, and predictive hiring algorithms, labor-market participation is increasingly filtered through automated evaluation mechanisms designed to optimize efficiency. In 2025, recruiters commonly reported receiving between 300 and 500 applications per position, making algorithmic screening an attractive solution for managing the overwhelming volume of applicants (Caldwell, 2025). However, the same systems that improve efficiency may also introduce structural barriers for workers already vulnerable to economic displacement.

This paper examines the relationship between automation-driven labor displacement and algorithmic hiring systems, arguing that automated screening technologies intensify labor market inequality among displaced workers, particularly those with non-traditional educational and career histories. Specifically, workers displaced by automation are likely to experience lower hiring progression outcomes, including callbacks, interview advancement, and job offers, when evaluated through algorithmic systems rather than human-led hiring processes. This penalty disproportionately affects individuals whose resumes reflect employment gaps, occupational transitions, retraining pathways, or alternative credential structures that deviate from historically “successful” hiring profiles embedded within machine-learning models.

By integrating labor-displacement theory with research on algorithmic hiring bias, this paper contributes to existing literature by reframing hiring algorithms as labor-market institutions rather than neutral technological tools. While prior scholarship has extensively examined automation’s impact on employment and separately explored bias in artificial intelligence, comparatively little research has connected these phenomena into a unified framework. This paper argues that algorithmic hiring systems shape post-automation labor allocation by controlling access to reemployment opportunities, thereby influencing productivity growth, labor mobility, and economic inequality at a macroeconomic level.

Automation and Labor Displacement

Substantial empirical research demonstrates that automation disproportionately displaces workers concentrated in routine-intensive occupations. Frey and Osborne (2017) estimated that approximately 47% of U.S. jobs face a significant risk of automation, particularly positions involving predictable, codifiable tasks. Similarly, Acemoglu and Restrepo (2020) found that the adoption of industrial robotics reduced both employment and wages in highly exposed labor markets. Manufacturing illustrates this transition clearly. Between 2015 and 2024, total U.S. manufacturing employment declined from approximately 15.3 million workers to 12.8 million workers while automation adoption accelerated across logistics, production, and warehouse operations (U.S. Bureau of Labor Statistics [BLS], 2024). These displacement effects are not evenly distributed. Workers without traditional four-year degrees, older workers, and career transitioners face elevated adjustment costs because they are more likely to occupy automation-vulnerable roles while possessing fewer institutional advantages during reemployment (Bessen et al., 2022).

Importantly, automation-induced displacement increasingly produces non-linear employment trajectories rather than permanent labor-force exit. Workers affected by technological disruption often transition across sectors, participate in reskilling programs, accept temporary employment, or accumulate alternative credentials through certifications and vocational pathways. These adaptive responses reflect rational economic behavior in changing labor markets. However, they also generate employment histories that diverge from the stable, linear career patterns traditionally associated with employability. This distinction becomes particularly significant within algorithmic hiring systems trained to recognize historical hiring patterns.

Algorithmic Hiring Systems and Structural Bias

Algorithmic hiring systems now play a dominant role in modern recruitment infrastructure. Large employers increasingly rely on automated resume parsers, ranking systems, keyword filters, and predictive scoring models to reduce applicant pools before human review occurs. These systems evaluate candidates using variables associated with prior successful hires, including educational pedigree, employment continuity, tenure stability, credential progression, and firm prestige (Dennison, 2025). Although these metrics appear neutral, they embed implicit assumptions about what constitutes a “qualified” worker. Linear career advancement, uninterrupted employment, and traditional educational pathways are rewarded as indicators of competence and reliability, while deviations from these patterns are frequently penalized.

The economic implications of this structure are significant because displaced workers are precisely the individuals most likely to possess non-traditional career trajectories. Workers adapting to technological disruption often exhibit temporary unemployment, retraining periods, cross-sector transitions, and credential diversification. However, algorithmic systems may interpret these patterns as indicators of instability or reduced productivity. From a machine-learning perspective, resumes that deviate from historical patterns are often treated as higher-risk inputs unless contextual variables are explicitly modeled. Consequently, hiring systems may systematically exclude qualified candidates whose employment histories reflect structural economic change rather than individual deficiency.

Evidence of Algorithmic Hiring Bias

Amazon’s AI Recruiting Case:

Evidence of algorithmic bias in hiring systems has become increasingly well-documented. One of the most widely cited examples emerged from Amazon’s experimental AI recruiting tool, which was ultimately abandoned after researchers discovered that the system penalized resumes associated with women and downgraded applicants whose language deviated from historically male-dominated hiring data (Dastin, 2018). Because the model was trained on prior recruitment outcomes, it replicated existing institutional biases embedded within historical hiring decisions. This case demonstrates that algorithmic systems optimize based on historical precedent rather than objective future productivity. Importantly, the underlying bias did not originate with the algorithm itself; it originated in historical human hiring decisions. However, by encoding those decisions into an automated system, AI can reproduce and scale existing biases across thousands of applicants with greater speed, consistency, and opacity than individual human decision-makers. Thus, the concern is not that AI creates new bias, but that it institutionalizes and amplifies existing patterns unless explicitly designed to mitigate them.

Audit Studies and Employment Gaps:

Additional audit studies further reinforce concerns surrounding automated screening bias. Research conducted through large-scale hiring experiments consistently identifies callback penalties associated with employment gaps and unconventional career pathways. Applicants with interrupted work histories are significantly less likely to receive interview invitations, even when qualifications remain otherwise comparable (Kristal et al., 2022). Studies examining workers who accepted temporary or transitional employment during periods of unemployment similarly found reduced callback rates relative to candidates with uninterrupted employment histories. These findings are especially concerning in automation-era labor markets where employment interruptions increasingly result from exogenous technological restructuring rather than individual underperformance.

Regulatory and Institutional Concerns:

Regulatory institutions have also recognized the growing risks associated with AI-driven hiring technologies. The U.S. Equal Employment Opportunity Commission (EEOC) warned that algorithmic hiring systems may violate federal anti-discrimination law when seemingly neutral screening criteria generate disparate impacts across protected groups (EEOC, 2023). Simultaneously, AI adoption has outpaced the development of formal governance structures. According to McKinsey Global Institute (Yee et al., 2025), approximately 78% of organizations report using AI in some capacity, while relatively few organizations maintain comprehensive governance frameworks capable of auditing or regulating algorithmic decision-making systems. This imbalance creates substantial institutional vulnerability as automated systems increasingly shape access to the labor market without sufficient oversight, transparency, or accountability mechanisms.

Macroeconomic Consequences of Algorithmic Exclusion

The consequences of these hiring dynamics extend beyond individual job seekers. Algorithmic hiring systems influence aggregate labor allocation by shaping which workers successfully transition into expanding industries following technological disruption. Classical labor-market models assume that employers can accurately interpret productivity signals and efficiently match workers to available opportunities. Human evaluators, despite their imperfections, retain the ability to contextualize employment gaps, career changes, regional economic shocks, caregiving interruptions, or retraining efforts. Algorithmic systems generally lack this interpretive flexibility unless such contextual factors are deliberately encoded into model design. As a result, displaced workers may encounter systematic exclusion precisely when labor markets require adaptability and mobility.

This exclusion produces measurable macroeconomic inefficiencies. Economic theory predicts that technological innovation generates long-run productivity gains when labor efficiently reallocates from declining sectors to expanding industries. However, if algorithmic screening restricts access to emerging occupations, labor mobility slows, and unemployment due to mismatches increases. Productive workers remain underutilized not because skills are absent, but because evaluation systems fail to recognize alternative pathways through which those skills were acquired. This represents a labor misallocation problem in which productive capacity exists but is inefficiently matched to demand.

The effects extend into broader patterns of inequality and economic growth. Research on displacement consistently demonstrates that automation-related job loss produces long-term earnings scarring, reducing lifetime income trajectories for affected workers (Bessen et al., 2022). If hiring algorithms delay or prevent reentry into higher-growth occupations, these income losses intensify. Lower earnings suppress household consumption, particularly in regions already heavily exposed to automation-driven restructuring. Reduced consumption weakens local economic recovery and reinforces geographic inequality. Over time, exclusion from high-growth sectors may deepen intergenerational inequality as access to economic mobility becomes increasingly constrained by algorithmic systems trained on historical patterns of privilege and stability.

Counterarguments and Limitations

Nevertheless, important counterarguments complicate the claim that algorithmic hiring necessarily worsens labor-market inequality. Some scholars argue that AI-driven hiring systems may reduce certain forms of human bias by standardizing evaluation criteria and minimizing subjective decision-making. Human recruiters themselves frequently demonstrate discriminatory tendencies related to race, gender, age, educational prestige, or implicit stereotyping. Automated systems can, in some cases, improve consistency and identify qualified candidates who might otherwise be overlooked during manual review processes. Additionally, algorithmic screening substantially reduces hiring costs and processing time, enabling firms to evaluate far larger applicant pools than would otherwise be feasible.

These arguments carry legitimate weight. Human-led hiring is neither neutral nor universally equitable, and eliminating automation would likely prove impractical in modern large-scale labor markets. However, human bias does not eliminate the risks associated with algorithmic exclusion. Instead, the concern lies in the scale, opacity, and institutional authority of automated systems. Human biases tend to vary across individual evaluators, whereas algorithmic systems apply screening criteria systematically across thousands or millions of hiring decisions simultaneously. Furthermore, because many hiring algorithms operate as proprietary “black box” technologies, applicants often lack transparency regarding how decisions are made or how biases might be corrected. The issue, therefore, is not whether algorithmic systems should exist, but whether current systems adequately account for the structural disruptions produced by automation itself.

Policy and Institutional Reform

Addressing these challenges requires both institutional and policy reform. Firms should conduct regular bias audits evaluating outcomes for workers with non-traditional career paths, alternative credentials, and employment interruptions associated with technological displacement. Human review mechanisms should remain integrated into hiring pipelines for applicants exhibiting non-linear employment histories to ensure contextual interpretation rather than automatic penalization. Additionally, hiring-system validation should prioritize downstream measures such as long-term performance, retention, and productivity rather than merely replicating historical hiring patterns.

At the policy level, algorithmic hiring systems should be treated as high-impact labor-market infrastructure rather than discretionary human resources tools. Regulatory frameworks should require transparency regarding model inputs, mandate disparate-impact testing consistent with existing employment law, and establish governance standards for AI-driven employment systems. Workforce-development initiatives must also recognize that reskilling alone cannot address labor-market inequality if hiring systems structurally filter out workers pursuing non-traditional pathways. Effective labor policy, therefore, requires simultaneous attention to both skill development and equitable access to employment opportunities.

Conclusion

Ultimately, automation and algorithmic hiring interact to form a dual exclusion mechanism. The first stage displaces workers through technological substitution, while the second restricts reentry through biased screening processes. What might otherwise represent temporary labor-market disruption becomes persistent exclusion when hiring systems systematically penalize the very employment patterns created by structural economic change. Because hiring algorithms increasingly shape who gains access to emerging industries, they influence not only individual employment outcomes but also patterns of labor allocation, productivity growth, and income distribution across the broader economy.

The broader significance of this issue extends beyond ethics alone. Algorithmic hiring bias represents structural labor-market inefficiency that can weaken the long-term benefits of technological progress. Innovation-driven growth depends on the economy’s ability to reallocate talent efficiently into emerging sectors. When access to those sectors is constrained by systems optimized for historical patterns rather than future potential, productivity gains are unevenly distributed and labor mobility declines. The sustainability of technological progress, therefore, depends not only on continued innovation but on whether labor-market institutions evolve to ensure inclusive access to opportunity within an increasingly automated economy.

References

Acemoglu, D., & Restrepo, P. (2020). Robots and jobs: Evidence from US labor markets. Journal of Political Economy, 128(6), 2188–2244. https://doi.org/10.1086/705716 

Bessen, J., Goos, M., Salomons, A., & van den Berge, W. (2022). Automatic reaction—What happens to workers at firms that automate? The Review of Economics and Statistics, 107, 125 (2025). https://scholarship.law.bu.edu/faculty_scholarship/584

Caldwell, S. (2025, October 29). Recruiters are “drinking through a fire hose” of job applications, experts say, and AI is partially to blame. CNBC. https://www.cnbc.com/2025/10/29/recruiters-are-drinking-through-a-fire-hose-of-job-applications-experts-say.html 

Dastin, J. (2018, October 10). Insight—Amazon scraps secret AI recruiting tool that showed bias against women. Reuters. https://www.reuters.com/article/world/insight-amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK0AG/

Dennison, K. (2025, August 18). The impact of hiring assessments on job seekers in the hiring process. Forbes. https://www.forbes.com/sites/karadennison/2025/08/18/the-impact-of-hiring-assessments-on-job-seekers-in-the-hiring-process/ 

Frey, C.B., & Osborne, M.A. (2017, January). The future of employment: How susceptible are jobs to computerization? Technological Forecasting and Social Change, 114, 254–280. https://doi.org/10.1016/j.techfore.2016.08.019. (https://www.sciencedirect.com/science/article/pii/S0040162516302244) 

Kristal, A. S., Nicks, L., Gloor, J. L., & Hauser, O. P. (2022, December 5). Reducing discrimination against job seekers with and without employment gaps. Nature Human Behavior. https://pmc.ncbi.nlm.nih.gov/articles/PMC7614241/

U.S. Bureau of Labor Statistics. (2024, May). Occupational Employment and Wage Statistics (OEWS) tables. U.S. Bureau of Labor Statistics. https://www.bls.gov/oes/tables.htm

U.S. Equal Employment Opportunity Commission. (2023, January 31). Meeting of January 31, 2023—Navigating employment discrimination in AI and automated systems: A new civil rights frontier [Meeting transcript]. https://www.eeoc.gov/meetings/meeting-january-31-2023-navigating-employment-discrimination-ai-and-automated-systems-new/transcript

Yee, L., Madgavkar, A., Krivkovich, A., Smit, S., Chui, M., Ramirez, M. J., & Castresana, D. (2025, November 25). Agents, robots, and US: Skill partnerships in the age of AI. McKinsey Global Institute. https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-skill-partnerships-in-the-age-of-ai