Abstract

This paper investigates the deployment of predictive policing systems in major American cities, interrogating the profound tension between their stated goal of achieving “algorithmic justice” and their operational reality as engines of “digital discrimination.” While proponents champion these technologies as objective, data-driven tools capable of overcoming human bias and enhancing law enforcement efficiency, this analysis argues that their current implementation institutionalizes and amplifies historical racial biases. The paper traces the genealogy of predictive policing from its roots in Compstat to its modern, commercialized form, deconstructing the technical architecture that perpetuates discrimination. Central to this critique is the reliance on “dirty data”-historical police records tainted by racially skewed enforcement practices-which fuels runaway feedback loops that concentrate police presence in minority communities, regardless of actual crime rates. Through critical case studies of programs in Chicago and Los Angeles, the paper demonstrates a consistent pattern of unproven efficacy, racial disparity, and eventual discontinuation following independent audits. Furthermore, it presents a rigorous constitutional analysis, arguing that predictive policing challenges the Fourth Amendment’s requirement for articulable suspicion and the Fourteenth Amendment’s guarantee of equal protection. The paper concludes that narrow technical fixes, such as algorithmic audits and fairness-aware machine learning, are insufficient to resolve these fundamental flaws. A genuine pursuit of justice requires a paradigm shift: from punitive prediction to restorative social investment, leveraging data not for targeted enforcement but to address the systemic inequities that are the root causes of crime.

Keywords