Predictive policing and criminal law
Manuel A. Utset
Introduction
Crime prediction technologies are routinely used by police to make law enforcement decisions (Perry et al., 2013). Predictive policing, as this practice is known, makes use of machine predictions about the place and time when crimes are likely to occur and about the identity of potential offenders and victims. Proponents of the practice argue that these predictive technologies help increase the overall level of deterrence (Mohler et al., 2015), but the evidence on deterrence is still, at best, inconclusive (Bennett Moses and Chan, 2018). At the same time, there is growing evidence that predictive policing comes at a cost, including the potential for errors and biases that are difficult to detect and prevent (Ferguson, 2017).This chapter considers whether, in light of these issues, predictive policing can be justified as a viable deterrence tool, and if so, under what circumstances, and at what costs. It will focus in particular on the use of predictive policing tools as a deterrence in real-time (what I will refer to as ‘real time policing’).
Real-time policing, as I will use the term, takes place within a small temporal window, starting when an offender decides to commit a crime and ending when the crime is completed and the offender has fled the crime scene. During this crime window, an offender undertakes a set of actions that can trigger criminal liability not just for the underlying crime, but also for inchoate offenses and a set of other corollary crimes. The chapter argues that real-time policing helps increase the overall level of deterrence by increasing the aggregate expected sanctions from following through with a planned crime. The chapter also identifies an important commitment problem that can lead to systematic underdeterrence of crimes that create relatively small losses. A statement by the state that it will fully enforce crimes after the fact will not be credible whenever the costs of investigating these crimes are much higher than the losses they produce. The chapter argues that real-time policing is a way for the state to precommit to enforce these low-loss crimes.
Real-time policing provides two other deterrence and crime-prevention benefits. Confronting offenders during the small crime window allows police to send salient signals about the expected costs of violating the law.The chapter argues that these salient signals can help better deter offenders who are highly impatient or have weak self-control. They also help deter two other types of offenders: those who commit crimes ‘erroneously’ because they are mistaken about the true magnitude of expected sanctions; and offenders embarking on a series of crimes in which they will learn-by-doing. This chapter describes a number of contexts in which offenders would be underdeterred if society relies on standard deterrence levers, but in which real-time policing can help close the deterrence gap.
Real-time policing, as the concept is used in this chapter, refers specifically to the real-time interaction between police and offenders during the crime window described above. Predictive policing algorithms can play an important role in real-time policing, as can other technologies that provide police with timely, actionable intelligence and other information that enhances their situational awareness (Wilson,2019). But police services must also examine whether the utilization of real-time crime forecasts and intelligence sufficient to drive real-time policing can be justified given the social costs involved. Each of the important questions outlined above will be examined within this chapter, which is divided into four parts. Part I provides an overview of standard deterrence approaches, including punishment-focused and police-focused deterrence regimes. Part II describes the most common uses of predictive technologies in law enforcement. Part III identifies a number of contexts in which real-time policing can help to better deter offenders and to prevent undeterred offenders from committing crimes; and Part IV describes a number of limitations of relying on machine predictions to make policing decisions.
I: Crime prevention and law enforcement
This part provides a brief overview of crime deterrence theories. It begins by describing the rational offender assumption commonly made in deterrence models. It then compares the punishment-focused deterrence approach of conventional law and economics models with the policing-focused deterrence approach that is influential among criminologists and police departments. The last section unpacks the concepts of crime prevention, law enforcement and real-time policing.
