Bilel Benbouzid's latest paper in Big Data and Society offers a detailed examination of the content of predictive policing applications. Crime prediction machines are used by governments to shape the moral behavior of police. They serve not only to predict when and where crime is likely to occur, but also to regulate police work. They calculate equivalence ratios, distributing security across the territory based on multiple cost and social justice criteria. Tracing the origins of predictive policing in the Compstat system, the paper studies the shift from machines to explore intuitions (where police officers still have control over the machine) to applications removing the reflexive dimension of proactivity, thus turning prediction into the medium for “dosage” metrics of police work quantities. Finally, the article discusses how, driven by a critical movement denouncing the discriminatory biases of predictive machines, developers seek to develop techniques to audit training dataset and ways to calculate the reasonable amount of stop and frisk over the population.
RSS Feed
