Risk And Countermeasure Research Of Predictive Policing

As artificial intelligence continues to evolve,the efficiency and precision of policing decisions will enter a new and broader space.Over the past decade or so,a growing number of police departments around the world have used a model of policing that uses data and algorithms to guide their decisions: predictive policing.Predictive policing predicts the time and space where crimes are likely to occur,and can even predict potential perpetrators and victims.Based on the results of predictions,police can plan patrol routes rationally and efficiently and assign officers to high-risk areas with the aim of preventing crime from occurring through deterrence and intervention.With the widespread use of predictive policing,the problems arising from it have gradually been discovered,which has aroused the concern of many people and scholars.Therefore,more and more scholars have begun to conduct in-depth studies on the possible risks brought by predictive policing,but most of these studies lack comprehensiveness and systematization and are often limited to a certain aspect or several aspects,without forming a more complete understanding of the content of predictive policing risks,and without classifying and studying the risks of predictive policing.At present,there is a lack of research on predictive policing risks in China.Supported by risk management theory,this paper uses literature analysis and Delphi method according to the three processes of risk identification,risk assessment,risk response and monitoring,and concludes that there are four categories of predictive policing risks: the first category is data risk,including dirty data,data security issues,insufficient data,lack of data transparency,and untimely data update;the second category is algorithm risk,including algorithm bias,algorithm opacity,and prediction results,algorithm opacity,wrong predicted outcomes,and limited predicted crime types;the third category is implementation risk,including lack of oversight and accountability,legal risk,wrong feedback loops,police bias and violence,police resistance to predictive policing,and overconfidence in predictive policing;and the fourth category is effectiveness risk,including lack of proof of effectiveness,lack of continuous evaluation,and failure to address the root causes of crime.Based on the above findings,an assessment system of predictive policing risk indicators was constructed;then the weights of each risk indicator of predictive policing were calculated using network hierarchy analysis,and the importance of predictive policing risk indicators were ranked.Among them,the lack of effectiveness of predictive policing proved to be the most important risk indicator.The next most important risk indicators are the lack of continuous evaluation of predictive policing and the existence of legal aspects of risk,respectively;finally,some countermeasures are proposed for the risks of higher importance.The research in this paper plays a certain guiding role for the risk management of predictive policing,which is conducive to improving the scientific level of implementing predictive policing and avoiding major risks,and has a certain reference value for the future development of predictive policing in China.