The Manchester Police Department has implemented a predictive policing system developed by IBM, which is able to identify hot zones for crimes in the city, mainly robbery, burglary and theft (Bonnette 2016). On the basis of historical data on crime hotspots, which contains a variety of variables related to crime (location, day, weather conditions, etc.), the system predicts where future criminal events are likely to take place (Dearden 2017). The system has been reported to cause a reduction in criminal activity in the targeted areas, as well as an improvement in public confidence in the police (UCL 2012).
As is the case with other predictive policing tools, the system presents great risks to privacy as well as the reproduction of structural societal biases. It has been well documented that when algorithms train from historical data sets, they learn how to emulate racially biased policing (ibid).
