← Papers

Targeted job‑seeker strategies—tailored resumes, channel choice, proactive messaging and upskilling—substantially raise hiring chances on Chinese recruitment platforms, offsetting much algorithmic screening bias; the result highlights microlevel agency as a complement to platform reform and regulation.

Job Search Game Under an Algorithmic Black Box: Generation of Group Biases in Recruitment Platforms and Individual Adaptation Strategies

Aoxue Hu · April 15, 2026 · International journal of Asian social science research.

openalex correlational low evidence 7/10 relevance DOI Source PDF

The paper argues and presents empirical evidence that targeted individual strategies—resume optimization, channel selection, proactive communication, and skill enhancement—can substantially mitigate algorithmic screening biases on Chinese recruitment platforms and improve job search success rates.

With the popularization of digital recruitment platforms in the era of artificial intelligence, algorithmic screening has become a core and indispensable component of talent matching in the modern labor market. However, inherent algorithmic opacity and historical data biases tend to give rise to obvious group prejudices based on gender, educational background, age, and regional origin, thereby further exacerbating the structural inequalities that exist in the current employment market. Existing academic research focuses primarily on the macrolevel governance paths of algorithmic discrimination, with relatively insufficient in-depth exploration of the microlevel game logic of job seekers and the construction of systematic adaptation strategies. In this paper, mainstream recruitment algorithms are taken as the core research object; the multidimensional specific manifestations and internal generation mechanisms of group prejudices in algorithm screening are systematically investigated; and the complex interactive relationships among job seekers, recruitment platforms, and enterprises as well as realistic individual predicaments are analysed on the basis of the classic theory of incomplete information games, and a scientific four-in-one adaptation strategy system encompassing resume optimization, channel selection, proactive communication, and ability enhancement is constructed. An empirical study revealed that active and targeted individual adaptation can effectively avoid the negative impact of algorithmic bias and significantly improve the overall job search success rates of different groups while providing important microlevel references for platform algorithm optimization and the improvement of relevant regulatory policies. It holds important practical significance for promoting the coordinated and sustainable development of efficiency and fairness in the field of digital recruitment in China.

Summary

Main Finding

Algorithmic screening on mainstream digital recruitment platforms functions as an opaque "black box" that reproduces and amplifies historical and social biases (gender, education/institution, age, region). Framed as a dynamic incomplete-information game among job seekers, platforms, and employers, these biases create systematic group-level disadvantages. The paper proposes a practicable four-in-one individual adaptation strategy (resume optimization, channel selection, proactive communication, and ability enhancement) and presents empirical evidence that active, targeted individual adaptation materially improves candidates’ algorithmic screening outcomes while also offering microlevel guidance for platform redesign and policy interventions.

Key Points

Algorithmic black box: Recruitment algorithms are largely unobservable and unexplainable externally; opacity + historical training data and design choices embed bias.
Dominant bias dimensions:
- Gender: Women face lower preliminary pass rates in technical, management, and professional roles; algorithms incorporate “stability” signals (age, marital/child status) that disadvantage women of typical childbearing ages. Reported figures: women’s pass rate in tech/engineering ~28.7% lower than men; in finance/legal ~17.3% lower. Non-fertile women are flagged “high risk” 3.2× as often as men and 1.8× as often as fertile women (source: platform monitoring reports cited).
- Education/institution: Heavy reliance on prestige labels (e.g., “double first-class”, full‑time study) produces stratification. Reported figures: matching scores for graduates of top universities ~41.6% higher; preliminary pass rates ~2.3× higher; part‑time education pass rate ~42.7% of full‑time. Many employers (cited ~67.4%) set explicit/implicit college thresholds; ~59.2% use “full-time undergraduate” as a minimum.
- Age/Region: Algorithms exhibit thresholding (notably a de facto 35‑year cutoff); reported pass rate for >35 is ~31.8% of the 25–30 group, with even lower rates in competitive sectors.
Theoretical framing: Integrates algorithm discrimination theory (data-driven and design bias), incomplete-information game theory (job seekers’ strategic constraints under opacity), employment equity theory (normative benchmark), and the social history of technology (technology shaped by social structures).
Individual adaptation: Presents a systematic four-in-one strategy:
Resume optimization (keyword and format alignment with inferred algorithmic features)
Channel selection (target platforms and employer types with different algorithmic thresholds)
Proactive communication (direct outreach to recruiters to bypass automation)
Ability enhancement (skills/certifications to alter observable signal set) Empirical results suggest such strategies can significantly increase pass rates across disadvantaged groups.
Policy and platform implications: Microlevel individual strategies are necessary under imperfect governance, but they do not substitute for platform transparency, fairness constraints, third‑party audits, and regulatory oversight.

Data & Methods

Conceptual and theoretical methods: literature review and synthesis; construction of a unified framework combining four theoretical lenses; game-theoretic analysis of the incomplete-information job search game among job seekers, platforms, and employers.
Empirical approach (as described): analysis draws on simulation delivery experiments and aggregate monitoring/industry reports from major Chinese recruitment platforms and research institutes (examples cited: Research Report on algorithmic discrimination in China's labor market 2024; Zhilian recruitment monitoring; Liepin Big Data Research Institute; Boss platform monitoring). These sources supply comparative pass‑rate and matching‑score statistics used to document bias magnitudes.
Strategy evaluation: an empirical study (methodological details in the full paper) tests the effectiveness of the proposed individual adaptation system and reports significant improvements in screening outcomes across groups. The paper emphasizes practical operability rather than proposing a single statistical model; it uses observed platform metrics and controlled resume-delivery experiments to assess impacts.
Note on limits: The excerpted paper relies heavily on platform monitoring reports and simulation experiments rather than a single longitudinal administrative dataset; precise sample sizes, statistical models, and robustness checks should be consulted in the full text for detailed replication.

Implications for AI Economics

Efficiency–equity tradeoffs: Algorithmic screening improves search efficiency but can institutionalize unfairness; markets may appear more efficient at matching on algorithmic signals while worsening allocation by true productivity when signals correlate with group identity rather than ability.
Distributional consequences: Persistent algorithmic group bias can entrench labor‑market stratification (credential inflation, occupational segregation, age‑based exclusion), contributing to long‑run inequality in employment and earnings across demographic groups.
Strategic behavior and signaling: Widespread adoption of individual adaptation strategies (resume tailoring, credential accumulation) creates endogenous responses that may raise search costs, generate arms races in credentialing, and shift equilibrium signaling equilibria—potentially increasing inefficiencies and inequality (those with resources adapt more effectively).
Platform incentives and regulation: Platforms have incentives to optimize for client (employer) satisfaction and speed, which can conflict with fairness goals. Economic policy responses include transparency mandates, third‑party algorithmic audits, mandated fairness constraints or counterfactual testing, and incentives for platforms to internalize distributional externalities.
Labor market design: Regulators and platforms should consider hybrid mechanisms—combining automated pre-screening with human review, randomized audits, and targeted support (e.g., visibility boosts or subsidized outreach) for historically disadvantaged groups to mitigate feedback loops.
Research agenda: Quantitative work is needed to estimate macroeconomic impacts (employment rates, wage dispersion) of algorithmic bias, to model equilibrium effects of large-scale individual adaptation, and to design incentive-compatible algorithmic fairness interventions that align platform revenue motives with social welfare.

If you want, I can: - Extract the paper’s empirical tables/figures and key numerical results into a concise table; - Draft short policy recommendations tailored to regulators, platforms, or job seekers; - Identify empirical tests and datasets that would strengthen causal claims in follow-up research.

Assessment

Paper Typecorrelational Evidence Strengthlow — The paper makes causal-sounding claims that individual adaptation 'effectively avoids' algorithmic bias, but the abstract provides no information on random assignment, plausibly exogenous variation, or robust identification techniques; estimates are likely subject to self-selection, omitted-variable bias, and unobserved heterogeneity between adaptors and non-adaptors. Methods Rigormedium — Strengths include a formal theoretical framing (incomplete-information game theory) and an empirical component that appears to analyze real recruitment-platform data; however, the empirical identification strategy and robustness checks are not described, limiting confidence in the rigor of causal inference even if descriptive and modelling work may be well executed. SampleUses data from mainstream Chinese digital recruitment platforms and job seekers (resumes, application outcomes and/or survey/behavioral data on adaptation strategies); the abstract does not specify sample size, selection criteria, time period, sectors, or how adaptation behavior was measured (self-report vs observed). Themeslabor_markets inequality adoption IdentificationNot clearly described in the abstract; the paper frames the problem using incomplete-information game theory and reports an empirical comparison of job seekers who adopt proposed 'four-in-one' adaptation strategies versus those who do not, but no exogenous variation, randomization, or clear instrumental strategy is reported to isolate causal effects. GeneralizabilityContext limited to Chinese digital recruitment platforms and institutional/regulatory environment, Findings may not generalize across platforms with different screening algorithms or firms with varying HR practices, Likely applies to groups represented in the sample; sectoral, occupational and firm-size heterogeneity not detailed, Results may not hold in markets with different labor laws, norms, or anti-discrimination enforcement, If empirical estimates are observational, effects may not generalize beyond self-selecting job seekers who choose to adapt

Claims (9)

Claim	Direction	Confidence	Outcome	Details
With the popularization of digital recruitment platforms in the era of artificial intelligence, algorithmic screening has become a core and indispensable component of talent matching in the modern labor market. Adoption Rate	positive	high	adoption of algorithmic screening in recruitment	0.3
Inherent algorithmic opacity and historical data biases tend to give rise to obvious group prejudices based on gender, educational background, age, and regional origin, thereby further exacerbating the structural inequalities that exist in the current employment market. Inequality	negative	high	group prejudice / structural inequalities in employment	0.3
Existing academic research focuses primarily on macrolevel governance paths of algorithmic discrimination, with relatively insufficient in-depth exploration of the microlevel game logic of job seekers and the construction of systematic adaptation strategies. Governance And Regulation	null_result	high	focus of academic research (macro vs micro)	0.15
Mainstream recruitment algorithms are taken as the core research object and the multidimensional specific manifestations and internal generation mechanisms of group prejudices in algorithm screening are systematically investigated. Other	null_result	high	manifestations and mechanisms of algorithmic group prejudice	0.15
The paper analyses the complex interactive relationships among job seekers, recruitment platforms, and enterprises on the basis of the classic theory of incomplete information games. Other	null_result	high	interactive relationships (game-theoretic analysis)	0.15
A scientific four-in-one adaptation strategy system encompassing resume optimization, channel selection, proactive communication, and ability enhancement is constructed. Skill Acquisition	positive	high	adaptation strategy components	0.15
An empirical study revealed that active and targeted individual adaptation can effectively avoid the negative impact of algorithmic bias and significantly improve the overall job search success rates of different groups. Hiring	positive	high	job search success rates (ability of adaptation to mitigate algorithmic bias)	0.15
These individual adaptation strategies provide important microlevel references for platform algorithm optimization and the improvement of relevant regulatory policies. Governance And Regulation	positive	high	policy and platform optimization relevance	0.05
The work holds important practical significance for promoting the coordinated and sustainable development of efficiency and fairness in the field of digital recruitment in China. Organizational Efficiency	positive	high	efficiency and fairness in digital recruitment	0.05