Tradeoffs in Hotspot Predictive Policing
Hotspot policing, which sends disproportionate police resources to small geographic areas, has become the dominant theory of data-driven law enforcement. However, predictive policing systems use narrowly scoped data and narrowly defined objectives that may lead to misalignment of police patterns with the interests of high-crime, poor and minority communities. We propose to study how lack of information about the effects of hotspot policing on communities-beyond their effect on crime-has affected predictive analytics and police department policies. We further analyze how collecting data about community interests beyond crime and incorporating community interests beyond crime reduction could lead to different outcomes or suggest different policing strategies. We propose to analyze a formal utility-based model of hotspot predictive policing and calibrate the model to newly available real-world data. Our work should have implications for law enforcement policy and is a step towards understanding issues of equity in data-driven policing.