Increasing workloads and dwindling resources demand police forces strive to enhance the efficiency of their operations and asset management. Consequentially, certain policing philosophies, like community-policing become less and less feasible as they demand resources often beyond the police’s capacity. To mitigate this issue, an older strategy from the 1920s has remerged with the aim of assisting police in not only resource placement and management, but also crime prevention. Coined ‘predictive policing’ this strategy introduces a quantitative element, data analysis, combined with essential elements of tried-and-tested policing strategies to enable law enforcement to make statistical predictions about potential criminal activity and thus guide their decision making. Predictive methods allow police to work more proactively with their constrained resources and several techniques have already been introduced into the field by law enforcement across the United States as well as reviews of their effectiveness. While anticipating criminal activities may seem Utopian, or even prophetic, it is not without criticism. Before getting to that however, it’s important to trace the idea back to its roots and methods.
A common myth about predictive policing is that it is a recent development however this couldn’t be more wrong. Law enforcement and crime analysts have engaged in predictive policing since the beginning of policing itself. To some degree or another people have used historic information concerning crimes, the surroundings, time, and the criminal's behaviour to develop a logical "prediction" regarding the occurrence of the possible next crime. A technique that has only been enhanced by the current technological state and enables the possibility for real-time analytical results to be produced. Examples of existing predictive technology used to anticipate crime, like the Dutch Crime Anticipation System and the American ‘Predpol’ demonstrate the significant role predictive analytics can play in the fight against crime. However, methods for predictive crime is but one element of predictive policing.
Equally as important as predicting crime, is predicting offenders, perpetrators’ identities, and potential victims of crime. Within each category, there are a range of approaches that can be applied that range from conventional analytics to predictive analytics. These are influenced by the level of data required for the analysis as well as the complexity of the analysis technique itself. For example, in the case of crime prediction, a conventional method would be to map crime areas, known as hot spot identification. In the conventional form this is performed using historical crime data that would be considered to be of a basic level. On a predictive analytical level a wider pool of data is used to create advanced hot spot identification models and conduct risk terrain analysis. What can be seen here is the enhancement of approaches that would rely on conventional human judgements, towards those which incorporate that into actively researched mathematical models. This transition also reflects a key tent of predictive policing: the strategy does not replace policing methods but serves to enhance them.
Making predictions is only half of predictive policing; however, with the other focus being on intervention in crime. Intervention, according to Perry et al. (2013) can be divided into three categories that range in complexity. The first category, generic intervention, focuses on increasing police resources in areas at great risk of crime. The second category , crime-specific intervention also focuses on resource allocation but tailors it to those which would be required for intervening in specific-crime types. The final category, problem-specific intervention, focuses on addressing specific locations and factors that are driving crime risks in a region. Each intervention ideally reduces or solves crimes but requires rapid assessment in order to achieve timely intervention.
As mentioned in the beginning, predictive policing appears quite attractive; however, it is not without its pitfalls and criticisms. The most obvious criticism pertains to the data sets used to influence algorithms that in turn will drive police decision making. In the United States, several studies have shown how predictive policing tools can perpetuate systemic racism through the use of biased data. For example, in 2019 Los Angeles dismantled its LASER program after an internal audit revealed that there were significant problems with the program, in particular inconsistencies in how individuals were selected and kept in data systems. As the program operated with inconsistent data, officers who relied on the data to police certain districts and neighbourhoods may not have had as much of a reasonable cause to make arrests as previously thought. Additionally, those previous arrests must be unbiased to generate unbiased predictions. Another problem found with date sets in the context of the US was the presence of racial bias. As people of colour are arrested more frequently than white people for committing the same crime, this creates racially biased crime predictions in districts where more people of colour have been arrested. From this perspective, predictive policing rather than being revolutionary is tyrannical in reinforcing unjust systemic racism.
With respect to the benefits of predictive policing, there are mismatches identified in studies on the achievements of it methods. While the predictive models aim to reduce crime through efficient and effective policing strategies, there is little evaluation on the usage of these models to say this has been achieved. Similarly, methods for predicting potential victims and offenders alike yielded no promising result. This, coupled with the issues of bias data and transparency, call into question the true value predictive policing offers to law enforcement. While it may be an anachronism to think that policing can continue without predictive models, in their current states they do little more than to fulfil one’s fantasy of pareidolia and that reason alone is not enough to justify their use