Paper short abstract:

Predictive policing programs harness the data-infrastructural components of police records and arrest data to produce geospatial risk profiles that are utilized by police departments. How are these software programs changing how crime is understood as a data-generative encounter?

Paper long abstract:

In the world of government procurement, one of the most sought after urban data infrastructures is known as 'predictive policing.' Using arrest records and crime data, predictive policing programs utilize algorithms that create geospatial risk profiles for urban spaces. Cash-strapped municipalities have turned to these analytics in order to more efficiently manage their police departments' resources. The rise of predictive policing over the past several years has also been tied to public debates over the role of the police, especially in the context of police brutality that correlates with waxing racial and economic disparity in cities the world over. Although even the most staunch proponents recognize the limits to predictive policing, it remains a glimmer of technological hope for a more objective, data-driven, and, assumedly, just future of policing.

Here I report on ethnographic research with the programmers and product managers behind one predictive crime analytics software, HunchLab. I focus on these actors' aspirations for police reform, which most often involve new ways to collect data not only on criminals but on police behaviors themselves. Biometric and behavioral data on police officers allows not only for predictions to be made, but also prescriptions about the best course of action. I describe this shift from 'prediction' to 'prescription' not only as a rhetorical device to distinguish the HunchLab product on the market, but also as a mechanism to enroll new actors (the police officers) into the interactive, data-generative field that crime becomes from the perspective of the data analyst.