What question did this study set out to answer?

This paper aims to analyze algorithmic bias in AI systems used for admission and hiring, highlighting the risks of inequality.

May 10, 2026Open Access

Bias in AI Admission and Hiring

Key Points

This paper aims to analyze algorithmic bias in AI systems used for admission and hiring, highlighting the risks of inequality.
Examined technical sources of algorithmic bias through a critical synthesis methodology
Analyzed the intersection of deep learning architectures and regulatory frameworks like the EU AI Act
Proposed a governance framework focusing on Algorithmic Impact Assessments and auditing processes.
Identified significant challenges with fairness metrics and the introduction of risks from large language models.
Proposed a shift from Human-in-the-Loop to Human-in-Command architectures for better accountability.
Concluded the need for AI systems to support human oversight rather than act as gatekeepers.

Abstract

The integration of artificial intelligence into admission and hiring systems offers modernizing efficiency yet entails significant risks regarding the automation of inequality. This paper examines the technical sources of algorithmic bias, including proxy variables, feedback loops and data provenance, within educational and workforce pipelines. Employing a critical synthesis methodology, the study provides a theoretical analysis of the intersection of deep learning architectures, emerging Generative AI (GenAI) capabilities and evolving regulatory frameworks such as the EU AI Act and NYC Local Law 144. The analysis highlights that “fairness through unawareness” is technically insufficient and that the deployment of Large Language Models (LLM) introduces new risks of algorithmic monoculture and sociolinguistic homogenization. Addressing the mathematical impossibility of simultaneously satisfying competing fairness metrics, the paper proposes a responsible AI governance framework. This framework emphasizes Algorithmic Impact Assessments (AIA), continuous disparate impact auditing, and a transition from Human-in-the-Loop to Human-in-Command architectures. The study concludes that responsible implementation requires shifting the role of AI from a gatekeeper to a decision-support scout, ensuring that automated systems meet rigorous standards of accountability and symbiotic human oversight.

Bias in AI Admission and Hiring

Key Points

Abstract

Cite This Study