PaperlayerView exact revision on arXiv

Source-linked AI summary

Can Labor Markets Function in the Age of AI? The Evaluation Bottleneck in Hiring

Itai Ashlagi, Ramesh Johari, Jon Kleinberg, Anushka Murthy

arXiv:2609.30058v1cs.GTcs.AIcs.CYecon.TH

TL;DR

AI-assisted job-search tools can make applications easier to submit while making their materials less informative about fit. The paper models how firms’ screening changes as that information deteriorates and examines whether multistage hiring can restore evaluation opportunities. It finds that inexperienced-compatible applicants are especially exposed, while intermediate assessments can give them a path to screening.

  • Problem

    AI can ease applications while reducing how informative applicant materials are about fit, raising the question of how this affects firms’ screening and access to evaluation.

    Source: §1 · pp. 1–2

  • Method

    The paper considers how firms screen applicants with different experience and compatibility when materials are noisy, and analyzes multistage hiring with an intermediate assessment.

    Source: §6.2 · p. 18, §7 · p. 23

  • Results

    Less informative materials lead firms to rely more on experience, disproportionately harming inexperienced-compatible applicants; costly screening can fail, while multistage assessment can restore some applicants’ path to screening.

    Source: §1 · p. 2, §7 · pp. 22–23

  • Takeaways & Limitations

    Intermediate assessments can create additional opportunities to demonstrate fit and help restore evaluation and hiring opportunities otherwise lost as application signals deteriorate.

    Source: §7 · p. 23

    The assessment rule does not characterize the globally optimal order of assessment and screening when multiple applicants remain.

    Source: §6.2 · p. 19

Abstract

from arXiv · show

AI-assisted job-search tools have become increasingly popular by making it easier to find and apply to jobs. But by making it easier for applicants to generate and tailor application materials, they can also reduce how informative those materials are about applicant fit. We study this tradeoff in a hiring market where applicants differ in experience and latent match quality and firms use noisy application materials to decide whom to screen. We ask how AI affects downstream screening and hiring, and which applicants are most adversely affected. As application materials become less informative, a Bayesian firm rationally relies more heavily on coarse observables such as prior experience. Among the four applicant types defined by experience and compatibility for the job, inexperienced-compatible applicants are the most exposed: they lack observable experience and lose the individualized information that could distinguish them from other inexperienced candidates. When screening is costly, these changes can also generate inefficient screening failures in which firms screen no applicants or screen only experienced applicants. We then show that multistage hiring can arise as an endogenous firm response: a relatively inexpensive intermediate assessment allows firms to acquire new evidence of fit before costly full screening. This can restore screening opportunities that disappear under one-stage hiring and give inexperienced-compatible applicants a path to screening. Our results show how AI can shift the central friction in hiring from submitting applications to obtaining credible evaluation, creating entry barriers for high-fit workers without prior experience. Multistage hiring can endogenously arise in response, restoring evaluation opportunities that would otherwise disappear and helping preserve market functioning.

1 Introduction

AI can make applying easier while making applicant materials less useful for identifying good matches. The paper models how this shifts screening toward experience and studies how intermediate assessments can reopen opportunities for inexperienced-compatible applicants.

1 Introduction

  • AI-assisted tools ease applications but may leave workers and employers facing more applications with less useful materials.The paper situates this concern amid reports of high application volumes, similar materials, and limited responses, particularly for new labor-market entrants.Source: §1 · pp. 1–2
  • As application materials become less informative, firms rely more on experience-based priors, disadvantaging inexperienced-compatible applicants who lose evidence that could distinguish them.Applicants without experience need stronger favorable evidence to offset experienced applicants’ prior advantage.Source: §1 · p. 2
  • Costly private screening can reject applicants who are worth screening socially, producing either no screening or exclusion of all inexperienced-compatible applicants.The firm does not internalize workers’ surplus from a successful match, and these failures can arise systematically as materials become uninformative.Source: §1 · p. 2
  • A relatively inexpensive intermediate assessment can generate additional applicant-specific evidence before costly full evaluation.The paper identifies an intermediate-assessment region where applicants may advance after a favorable assessment despite being rejected under one-stage hiring.Source: §1 · p. 3
  • When that region is nonempty, firms strictly prefer multistage hiring, which can avoid one-stage screening failures and give inexperienced-compatible applicants a path to screening and hiring.At high AI saturation, the benefit extends to nearly all submitted inexperienced-compatible applicants.Source: §1 · p. 3

2 Related Literature

The paper connects its analysis to research on AI and labor-market signaling, directed and sequential search, and statistical discrimination. Its focus is how degraded applicant signals affect access to costly individualized evaluation across experience groups.

AI and labor-market signaling

  • Related work finds that generative AI can weaken applicant-generated signals and change the evidence employers use, while this paper studies access to costly individualized screening and multistage adaptation [22] [28].It contrasts its focus with Jungbauer, who studies disruption to assortative matching and redesign of a common screening signal.Source: §2 · p. 3

Directed and sequential search

  • The paper adds an information friction to online labor-market models: signals used to decide whom to screen can themselves become less informative.Arnosti et al. study a decentralized market with applicant application costs and employer screening costs.Source: §2 · p. 4
  • The one-stage baseline specializes Weitzman’s sequential inspection problem to hiring, while the multistage model permits partial information acquisition before full screening.The analysis asks how sequential evaluation changes as applicant-generated information deteriorates and experience creates heterogeneous priors about fit.Source: §2 · p. 4

Statistical discrimination and entry barriers

  • As applicant materials become less informative, experience plays the role of an observable characteristic correlated with expected productivity, rationally receiving greater weight in screening.The paper’s distinction from employer-learning models is its focus on whether applicants receive costly information acquisition needed to overcome an unfavorable prior.Source: §2 · p. 4
  • The paper’s static analysis differs from work on temporary two-stage hiring that shifts long-run market equilibrium, focusing instead on generating applicant-specific information before full screening [32].Reduced access to interviews and jobs may also limit inexperienced workers’ ability to accumulate experience firms rely on.Source: §2 · p. 4

3 Model

The model studies a firm screening applicants who differ in observable experience and latent match quality. AI saturation affects application costs, signal informativeness, and effective screening costs, while firms use observed experience and noisy materials to choose whom to screen.

3 Model

  • AI saturation q is modeled as affecting application costs, application-material noise, and the firm’s effective screening cost.As AI saturation increases, application costs fall and effective screening costs rise; submitted materials become less individually informative.Source: §3 · pp. 4, 6
  • Applicants have observable experience and latent match quality, with compatibility determined by match quality and four realized types defined by experience and compatibility.Experienced applicants have higher expected match quality than inexperienced applicants.Source: §3 · p. 5
  • Applicants observe their experience and application-cost scaling, but not match quality, and decide whether to apply before the firm observes experience and noisy materials.Application-cost scalings capture differences in the time, effort, or opportunity costs of preparing and submitting applications.Source: §3 · pp. 5, 7
  • The firm forms posterior compatibility beliefs from experience and materials, pays to screen applicants, and learns match quality through screening.It hires the first compatible applicant screened, if any; compatible matches generate payoffs for both the firm and applicant.Source: §3 · pp. 4, 6, 7
  • The model assumes both experience groups occur and apply with positive probability, and that higher AI saturation strictly increases signal-noise variance.AI may also raise effective screening costs because generated materials can become more homogeneous or less informative.Source: §3 · p. 6

4 Bayesian Reweighting and Firm Screening

The firm updates experience-group priors using noisy application materials, then screens applicants in descending order of posterior compatibility. As AI saturation rises, Bayesian updating places less weight on individualized materials and more on experience.

4 Bayesian Reweighting and Firm Screening

  • The posterior-mean score combines the materials signal with the experience-group prior, with the materials weight κ(q) and an experience component θ(q).The firm’s posterior compatibility is ordered in the same way as these scores.Source: §4 · pp. 7–8
  • As AI saturation rises, the firm places less weight on the materials signal and more on the experience-based prior.Consequently, the materials evidence needed to offset the prior difference between experienced and inexperienced applicants increases.Source: §4 · p. 8
  • The firm ranks applicants by posterior compatibility and screens sequentially until it finds a compatible applicant or no remaining applicant meets its screening cutoff.The cutoff reflects the comparison between the expected private value of screening and its cost.Source: §4 · pp. 8–9
  • The private posterior-compatibility cutoff rises with AI saturation when effective screening costs rise.The screening policy can equivalently be represented by a cutoff on posterior-mean scores.Source: §4 · pp. 8–9

5 Impact on Workforce Entry and Market Failure

As AI saturation rises, inexperienced-compatible applicants lose the individualized evidence that could offset their weaker experience-based prior. Costly screening can then be privately unprofitable even when socially valuable, producing failures that become increasingly likely as application materials lose informativeness.

5.1 Shortlisting Exposure

  • Inexperienced-compatible applicants have the largest negative change in conditional mean score as AI saturation rises, because weaker experience-based priors leave them without a compensating advantage.Incompatible applicants’ mean scores rise as individualized information loses weight, while compatible applicants’ scores fall.Source: §5.1 · p. 10
  • Noisier materials reduce compatible applicants’ chance of clearing a fixed threshold, while inexperienced applicants face a higher relative threshold whose gap from experienced applicants grows.Inexperienced-compatible applicants are exposed to both disadvantages.Source: §5.1 · p. 12

5.2 Market Failure

  • The firm’s private screening cutoff exceeds the social cutoff, leaving applicants whom a planner would screen but the firm rejects; higher effective screening costs expand this range.The firm does not internalize workers’ surplus from a successful match.Source: §5.2 · pp. 12–13
  • As materials become uninformative, screening failures can occur with probability approaching one: firms may screen nobody or continue screening experienced applicants while excluding inexperienced-compatible candidates.Posterior beliefs collapse toward experience-group priors, so realized compatibility becomes harder to distinguish within each group.Source: §5.2 · pp. 14, 16

6 Multistage Hiring as a Response to AI-Induced Entry Barriers

A lower-cost intermediate assessment can let firms gather new evidence of fit before costly full screening. The firm may benefit strictly, while inexperienced-compatible applicants gain a path to screening that expands as AI saturation rises.

6 Multistage Hiring as a Response to AI-Induced Entry Barriers

  • The response studied is a cheaper intermediate assessment, such as a work sample or live skill test, before costly full screening.The assessment supplies additional applicant-specific information; submitted applications can be retained without assessment cost.Source: §6 · p. 16, §6.1 · pp. 16–17

6.1 Multistage Screening

  • An intermediate assessment refines the compatibility posterior before full screening, and its information value can justify paying the assessment cost.Before accounting for assessment cost, posterior refinement cannot lower the firm’s expected continuation payoff.Source: §6.1 · pp. 17–18

6.2 A Binary Assessment and Multistage Screening

  • When an applicant is the firm’s only remaining option, a favorable binary assessment can lead to full screening, while an unfavorable result leads to rejection.The binary assessment is more likely to pass compatible than incompatible applicants within each experience group.Source: §6.2 · pp. 18–19
  • A multistage policy can strictly improve on one-stage hiring when a rejected applicant remains worth assessing, giving the firm positive ex ante value from the option.A feasible policy can first replicate one-stage hiring, then assess such an applicant if no hire occurs.Source: §6.2 · p. 19

6.3 Implications for Screening Failure

  • Multistage hiring can mitigate one-stage failures in which no applicant is screened or no inexperienced-compatible applicant is screened.An inexperienced-compatible applicant rejected under one-stage hiring can become worth assessing if reached while the vacancy remains open.Source: §6.3 · pp. 20–21

6.4 Implications for Inexperienced-Compatible Applicants

  • In the high-AI limit, almost all inexperienced-compatible applicants can gain a positive hiring-probability path through multistage hiring.This holds when their inexperienced-group prior lies above the assessment cutoff and below the one-stage screening cutoff.Source: §6.4 · p. 22

7 Conclusion

In the model, less informative application materials shift firms toward experience-based priors, especially disadvantaging inexperienced-compatible applicants. Costly screening can produce failures, while multistage hiring can reopen evaluation and hiring opportunities.

7 Conclusion

  • Less informative application materials lead firms to rely more heavily on applicants’ experience.The change affects what firms can infer from applications, not only how many applications workers submit.Source: §7 · p. 22
  • Inexperienced-compatible applicants are especially harmed when they lose individualized evidence that could distinguish them from inexperienced peers.Because firms do not internalize workers’ match surplus, they may reject socially worthwhile applicants; in very noisy environments, they may screen no one or only experienced applicants.Source: §7 · p. 23
  • A sufficiently informative, inexpensive intermediate assessment can increase firm payoffs and give some one-stage-rejected inexperienced-compatible applicants a positive-probability path to employment.As AI saturation becomes sufficiently high, the benefit can extend to almost all such applicants.Source: §7 · p. 23
  • As application signals deteriorate, multistage hiring can help restore evaluation and hiring opportunities that would otherwise disappear.The conclusion emphasizes access to credible evaluation for high-potential workers without conventional experience.Source: §7 · p. 23

A Proofs for Section 4

The proofs establish the firm’s screening order and stopping rule: applicants are considered by posterior compatibility, and screening stops when further inspection is not worthwhile.

A Proofs for Section 4

  • The proof equates posterior compatibility ranking with ranking by the posterior-mean score used to order applicants.The displayed equivalence holds for any submitted applicants.Source: A Proofs for Section 4 · p. 28
  • Conditional on the firm’s observed information, applicants’ compatibility realizations are independent.This follows from independence of applicant primitives conditional on the observed information.Source: A Proofs for Section 4 · p. 28
  • Pairwise interchange shows that the firm can order applicants by decreasing posterior compatibility and screen sequentially until no remaining applicant is worth screening.The proof uses that the order does not change the probability of finding a compatible applicant but can change expected screening costs.Source: A Proofs for Section 4 · pp. 28–29

B Proofs for Section 5

The proofs characterize a gap between the firm’s private screening cutoff and the planner’s social cutoff. Under stated conditions, that gap yields no-hire or inexperienced-compatible screening failures, which become likely as application materials grow noisy.

B Proofs for Section 5

  • The proof’s ordering of screening-value changes across applicant types is ΔS̄01(q, eq) < ΔS̄11(q, eq) < 0 < ΔS̄00(q, eq) < ΔS̄10(q, eq).The inequality preserves the experience-compatibility group labels used in the passage.Source: B Proofs for Section 5 · p. 31
  • Applicants with posterior compatibility between the planner’s cutoff and the firm’s cutoff are socially worth screening but excluded from the firm’s list.The proof states the interval as pP(q) ≤ pi(q) < pF(q).Source: B Proofs for Section 5 · p. 33
  • If every submitted applicant falls below the firm’s cutoff while at least one exceeds the planner’s cutoff, the firm screens no one although the planner screens someone.The proof establishes this no-hire screening failure under the specified cutoff condition.Source: B Proofs for Section 5 · p. 33
  • Under the stated cutoff conditions, submitted inexperienced-compatible applicants fall below the firm’s cutoff, while at least one is socially worth screening.The proof concludes that at least one such applicant is socially worth screening but excluded by the firm.Source: B Proofs for Section 5 · p. 33
  • For a fixed finite pool under the high-noise conditions, the no-hire failure or exclusion of every submitted inexperienced-compatible applicant occurs with probability approaching one.The respective limiting claims are stated for the two screening failures.Source: B Proofs for Section 5 · p. 35

C Proofs and Supporting Results for Section 6

The supporting results characterize when an intermediate assessment is worthwhile and establish that multistage hiring can outperform one-stage screening. They also show how assessment can reopen a screening path for inexperienced-compatible applicants excluded under one-stage hiring.

C Proofs and Supporting Results for Section 6

Loading 2609.30058v1…