Fairness-aware Machine Learning for Public Sector Resource Allocation

Abstract

Public sector organizations increasingly employ machine learning to guide the allocation of limited resources across populations and regions. While such systems offer gains in efficiency and scale, they also risk reinforcing historical inequities embedded in administrative data. This study presents a fairness-aware machine learning framework for public sector resource allocation that balances predictive performance with equitable outcomes. The proposed approach integrates fairness constraints, interpretable modeling, and post allocation auditing within a unified decision support pipeline. Empirical evaluation demonstrates that fairness-aware optimization can substantially reduce allocation disparities while preserving operational effectiveness. The results provide practical guidance for deploying trustworthy machine learning systems in high impact public decision contexts.