(Pre-)crime and Punishment

Do models that claim to be able to predict crime open the door to smarter policing, or just reduced policing?

“A city is an ecological system” – a true urban jungle, in which human behaviour can be viewed through a thoroughly biological lens, its stable patterns open to tracking and even anticipation. So says University of Miami Professor Chris Cosner, one of a growing number of academics who believe that the unlikely fusion of mathematics and ecology can lend valuable insights to the socioeconomic problems of our own natural systems.

Take crime, for instance. In a recent paper published in the SIAM Journal on Mathematical Analysis, Dr Cosner, with colleagues Stephen Cantrell, also of the University of Miami, and Raúl Manásevich, of the Universidad de Chile, Santiago, used techniques from mathematical ecology to predict patterns of residential burglary. Their tools? Differential equations, combined with a smattering of criminology and an understanding of how other organisms exploit their natural resources.

The theory that underlies this research purports that, when it comes to geography, crime is anything but random. We all recognise the existence of a ‘bad part of town’, and, if we live there, know better than to opt for the unlit alleyway when selecting our best route home at night. This is because we are subtly aware of what in criminology is known as the “broken-window effect”: the propensity for areas in which crime has already occurred to suffer from repeat offences.

“Burglars find places more attractive if they’ve broken into them before… or if their friends have broken into them”, explains Dr Cosner. Using this “attractiveness value” as a foundation, he and his colleagues built on existing work to develop a robust mathematical model that connects the geographical characteristics of an area to its likely patterns of burglary. Thus, by tracking patterns of attractiveness and accounting for factors such as a neighbourhood’s demographics and economics, they were able to isolate acutely vulnerable “hotspots”.

Such predictive models have obvious appeal at a time when public spending cuts have led to calls for smarter and more efficient policing from increasingly stretched forces. A prototype system, based on work carried out at UCLA, has been in operation in Los Angeles since November 2011, where a computerised “forecasting tool” capable of identifying emerging hotspots has contributed to a 33% reduction in residential burglary. The increased mathematical rigour of the more recent study has the potential to drive further improvements in the effectiveness of such systems.

However, efficiency may come at a price. While mathematical models might successfully predict general patterns of crime, they cannot foresee specific incidents, and some believe that increased reliance on technology comes at the cost of good, old-fashioned policing. “The fear with technologies that claim to be able to ‘predict’ crime is that they will miss serious crimes and potentially put members of the public at risk”, says Emma Carr, Deputy Director of the UK-based think tank Big Brother Watch, which also questions the threat posed to civil liberties by technologies designed to predictively tackle crime.

Others challenge the ability of mathematics to accurately model such complex social systems. Agent-based modelling, an alternative technology in which computer simulations imitate the behaviour of virtual people in computer-generated cities, has already produced promising results, and is heralded by its architects as a tool uniquely capable of dealing with the complexities of the real world.

After all, if a city really is an ecological system, it is an extremely complicated one. While swelling crime figures and shrinking tax revenues make it tempting to believe in our ability to predict and proactively prevent crime, only time and much more research will tell whether or not we really can.

Image: via Pacific Standard

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *