A challenging environment for police forces
Police forces are more than capable of incorporating new technologies and systems in specific ways, limiting their use to particular purposes and setting standards of performance and accountability. A tapestry of rules and regulations already surrounds the use of digital cameras, smartphones, personal computers, the compilation of case files and the processing of digital information, among other technologies, in a variety of policing contexts. Therefore we should not immediately discount the ability of police services, parliament, government bodies and civil society to work together to build a sensible architecture around predictive and AI policing systems. However, a variety of external pressures can cause police forces to eschew historical antecedents. Projections suggest that police services across the UK will be hundreds of millions of pounds in deficit by 2021 (National Audit Office, 2016). Years of austerity mean that the wider criminal justice system is faced with an “avalanche of problems” from the inability' to gather and present crucial evidence within shorter timeframes to a growing shortage of solicitors and socio-economic barriers to legal aid (Law Society, 2019: 14). Complicating the landscape further is the fact that a broader shift from policing as a reactive, crime-fighting role to more of a victim-focused, risk-based, preventative role is taking place in various crime areas. Addressing these problems through cost-effective and scalable technologies carries a particular attraction. Police forces may even consider the procurement of such technologies to be necessary to carry out their basic functions and to maintain the rule of law with the limited resources available to them (Ibid).
Not only are police services attracted to the promise of innovative predictive and AI technologies but they also have broad appeal across the political spectrum. British governments have provided millions of pounds of funding to police services to implement digital transformation projects in recent years. Public support is also palpable. The promise of drastically reduced crime rates, increased public safety, better value for money and smart-on-crime policies ticks many boxes. These factors arguably create quite strong incentives for police services to develop and employ digital technologies within remarkably short timeframes. Their allure is such that some police services have even procured them with little or no competitive bidding or evidence that they work in practice (Stevenson, 2018; Valentine, 2019). The culture of positive discourse around predictive and AI technologies in policy-making circles allied to the financial and external pressures to use them to ‘fix’ policing has created a challenging, and potentially toxic, environment. Stakeholders and the public should be cognisant of these incentives and pressures as they proceed, for they may serve to drown out the critics. In this vein of thought, we endeavour in the next two sections to introduce the reader to predictive policing and AI as two distinct, yet overlapping, features of modern policing, each with their own unique processes and issues. Generating greater clarity around what they represent is surely a step in the right direction.
What is predictive policing?
Predictive policing is a relatively new concept that involves the application of analytical techniques to data for the purposes of generating statistical predictions about events so that something can be done about them in advance. It is tempting to discuss factors such as crime and harm prevention in a description of predictive policing, but it arguably has more general qualities. Predictions can concern offenders, victims, police behaviour and organisational efficiencies among an array of other applications. Calculating the risk that X may lead to Y in any policing context and doing something about X or У could be construed as predictive policing. Commentators have even pointed out that predictive policing is a misnomer since a prediction is usually subjective, relies upon intuition, often involves a right or wrong outcome and is nonreproducible, whereas a process that is objective, scientific and reproducible and comes with an associated probability' is better described as a ‘forecast’ (Perry et al., 2013). Although much of what happens in this field would be better described as forecasting, it is predictive policing rather than forecast policing that has become the more popular term (Ratcliffe, 2019). What we do know is that there should be two key parts to predictive policing in practice: 1) generating a prediction, and 2) carrying out some type of policing intervention, investigation or activity' as a result (Perry et al., 2013).