Algorithmic Bias in Municipal Decision-Making: A Case Study of AI Deployment in Urban Public Service

Abstract

Artificial intelligence systems are increasingly embedded in municipal governance, influencing how cities allocate resources, enforce regulations, and deliver public services. This article investigates the manifestation and consequences of algorithmic bias in three urban public service domains: predictive policing, welfare benefit eligibility screening, and housing inspection prioritization. Drawing on a multi-city case study methodology spanning twelve North American and European municipalities from 2018 to 2024, we identify four recurring structural conditions that amplify bias in municipal AI: historically skewed training data, opaque procurement processes, the absence of community input in system design, and inadequate post-deployment auditing. Our findings demonstrate that algorithmic outputs often replicate and entrench pre-existing socioeconomic and racial inequalities rather than remediate them. We propose a Municipal AI Accountability Framework (MAAF) encompassing algorithmic impact assessments, participatory design requirements, mandatory transparency disclosures, and independent third-party auditing. The study concludes with policy recommendations directed at legislators, city administrators, and technology vendors.