AI-powered Talent Analytics for Inclusive Hiring Practices

Thesis Abstract

In contemporary labor markets, persistent biases in recruitment processes hinder workforce diversity and productivity, despite advances in human capital management. This study investigates how AI-powered talent analytics can enhance inclusive hiring practices by mitigating selection bias, improving candidate fit assessment, and fostering equitable decision-making across stages of the recruitment lifecycle. The aim is to develop and empirically validate a framework that integrates algorithmic screening, blind resume techniques, and fairness-aware predictive modeling to support inclusive hiring. The objectives are (i) to design an AI-enabled talent analytics pipeline that incorporates blind feature engineering and fairness constraints; (ii) to evaluate the impact of the pipeline on reducing demographic parity and predictive bias across gender, ethnicity, and disability status; (iii) to examine the relationship between analytics-driven hiring decisions and subsequent job performance, retention, and employee engagement; (iv) to identify organizational factors that influence adoption, trust, and ethical use of AI in recruitment; and (v) to formulate practical guidelines for implementing fairness-aware AI tools in hiring processes. The study employs a mixed-methods design combining quantitative evaluation of predictive models with qualitative insights from HR practitioners. A multi-site organizational sample comprising 12 mid-to-large firms across technology, manufacturing, and professional services sectors will be used, with an estimated population of 8,500 applicants annually. A stratified random sample of 1,200 applicants’ anonymized data will be analyzed, alongside 240 hires and 120 new hires observed over a 12-month follow-up period. Quantitative analysis will utilize machine learning techniques, including logistic regression, gradient boosting, and fairness-aware algorithms (e.g., equalized odds constraints) to predict job performance indicators while controlling for protected attributes. Model performance will be assessed via AUC, calibration plots, and fairness metrics such as disparate impact ratio and equal opportunity difference. Multilevel modeling will examine organizational context effects on hiring outcomes. Qualitative data will be collected through semi-structured interviews with 20 HR managers and 30 interviewees from underrepresented groups to explore perceived fairness, trust in AI decisions, and on-the-ground challenges. Thematic analysis will identify recurrent patterns, contrasted with theoretical expectations. Key expected findings include (a) a demonstrable improvement in candidate shortlisting accuracy with comparable or reduced bias when using the fairness-aware analytics pipeline; (b) evidence that removing identifiable demographic signals during initial screening reduces disparate impact without sacrificing predictive validity for job performance; (c) positive associations between AI-assisted inclusive hiring and subsequent performance, tenure, and engagement metrics; (d) critical organizational enablers and barriers to AI adoption, including governance structures, data quality, leadership support, and ethical training, with variations across industry contexts; and (e) a validated conceptual model linking technology, process redesign, and workforce diversity outcomes. The study contributes to knowledge by advancing the theory and practice of AI in HRM under the lens of inclusivity, offering a robust, replicable framework for fairness-aware talent analytics that can be generalized across sectors. It integrates theories of algorithmic fairness, signaling theory in recruitment, and technology-organization-environment (TOE) framework to elucidate how technical design choices interact with organizational norms to influence outcomes. Practical contributions include detailed technical specifications for implementing blind screening and fairness constraints, a set of performance and bias metrics, and a step-by-step implementation roadmap for organizations seeking to operationalize inclusive AI-driven hiring. The principal conclusion anticipates that ethically designed AI-powered talent analytics can enhance hiring equity while maintaining or improving predictive validity, provided that strong governance, continuous monitoring, and transparent communication with stakeholders are established. Recommendations emphasize establishing fairness-by-design protocols, ongoing audit processes for bias detection, stakeholder engagement strategies, and capacity-building initiatives for HR professionals to interpret and contest AI-driven decisions.