ABOUT BOOK

Predictive policing is a tool-based process that is increasingly framed in terms of AI. The aim is to identify patterns from large datasets that can help officers to focus their work and render police work more efficient. Such practices stretch from the generation of geospatial predictions to organize police patrol, to predicting physiognomic features from the DNA profile of an unidentified owner to guide investigations. Predictions are generated in a complex process involving humans, datasets and algorithms. This chapter follows the lifecycle of algorithms to illustrate these socio-technical clusters. It traces the imaginaries about the data expected to work best for predictions. It describes the collaborative practices of data generation, cleaning and storage, as well as the process of using this data for training algorithms. The chapter places these insights in a critical discussion about predictions as a socio-technical practice that performs its own politics.