Differential Privacy-Enhanced Federated Learning for Fair and Transparent Workforce Assessment

Abstract

The increasing adoption of workforce analytics has made organizations more reliant on data-driven decision-making for employee performance assessment, career progression, and organizational development. However, centralized workforce data collection poses critical risks to employee privacy, fairness, and transparency. To address these challenges, this research explores a novel framework that integrates Differential Privacy (DP) within Federated Learning (FL) to enable secure, equitable, and explainable workforce assessment. By combining the decentralized collaborative training of FL with the rigorous privacy guarantees of DP, the proposed model safeguards sensitive employee information while minimizing bias in performance metrics. Experiments conducted on simulated workforce performance datasets reveal that the DP-enhanced FL framework achieves competitive accuracy compared to centralized models while significantly improving privacy protection and fairness indices. The findings demonstrate that DP-FL can serve as a scalable and ethical solution for workforce analytics, balancing organizational utility with employee trust.