AI Recruitment Bias Governance through Multi-Case Comparative Study: From the Infeasibility of “Zero Bias” to Auditable Compliance and Engineering Practices

Artificial intelligence is revolutionizing recruitment, with 83% of employers using automated screening systems [14]. While boosting efficiency, AI introduces algorithmic bias and transparency issues, as seen in cases involving Amazon and HireVue [14]. Through case studies of Harver, Eightfold AI, HireVue, and LinkedIn, this study finds bias stems from the interaction of data, algorithms, and human interpretation [2;12]. Bias in AI recruitment systems does not arise from isolated technical flaws but from a structural coupling between social inequality and computational optimization. Historical labor market inequalities shape training data distributions; these distributions are then formalized through algorithmic objective functions (e.g., predictive accuracy or retention likelihood), which systematically privilege historically dominant groups. I propose the Auditable Fairness Framework (AFF)—based on Auditability, Engineering, Control, and Remediation—shifting the goal from unachievable “zero bias” to establishing detectable, explainable, and correctable governance.