Diagnosing Algorithmic Bias in AI-Powered Hiring: Toward a Fairness-Aware Framework

Artificial intelligence (AI) is gaining wider usage in hiring from resume screening through candidate evaluation. Though such systems hold out the promise of efficiency, they also understandably generate recurrent doubts about fairness and bias. This work addresses these doubts by presenting a diagnostic framework for detecting bias in AI-powered hiring. The framework combines three stages: (1) descriptive analysis of demographic distributions, (2) statistical testing of group outcome disparities, and (3) model-level fairness metrics, including selection rates, true positive rates, and false positive rates. With three extensively researched datasets: Adult Income, Job Salary, and COMPAS, this work demonstrates that gendered differences persist. Results demonstrate steady salary outcome gaps and differential error rates across demographic groups despite standardized preprocessing. The findings suggest that automated hiring tools without systematic fairness evaluation may reinforce existing workplace inequalities. Proposed here is a structured and reproducible method to detect algorithmic bias that will help both researchers and users promote fairness and accountability in hiring systems.