Deterrence or prevention? The challenges of algorithmic crime prediction in urban spaces

Abstract

Predictive policing emerged as a promise to prevent crime before it occurs. Despite the waning hype, predictive policing remains an active practice, especially in urban contexts. Drawing on ethnographic research on a predictive policing system of a German state police force, this paper examines the territorial and organizational conditions shaping the implementation and effectiveness of algorithmic crime forecasts. We develop a comparative analysis of their application in both urban and rural contexts. Our findings show that predictive policing is technically and conceptually tailored to urban areas, with limited potential for effective implementation in rural areas. Its practical implementation in urban areas, however, is often hindered by resource constraints, organizational hurdles, and the short-term logic of deterrence. As a result, the effectiveness of predictions is often limited even in the urban settings for which they were designed. We contrast this deterrence-oriented model with emerging preventive approaches based on micro-segmented spatial analysis, which seek to transform the conditions enabling crime rather than merely forecasting its occurrence. By distinguishing between deterrence and prevention, we argue that sustainable algorithmic crime prevention in urban spaces requires reconfiguring predictive technologies as tools for collaborative, long-term interventions rather than short-term operational targeting.