Just this year we’ve seen open data give rise to recreations of Denmark in Minecraft, the ability to compare cities at the same scale and also collections of geo-mapped tweets and traffic lights. But what about a practical application for all of that info, one that has a more tangible benefit to society, like, say, crime prediction? That’s what the University of Trento in Italy had in mind with its “Once Upon a Crime” study. The researchers coupled freely available demographic and mobile phone data with real crime data to forecast where in London an infraction might occur.
In the 2002 film, Minority Report, the police apprehend criminals by predicting that they are about to commit a crime. Many observers have drawn parallels between this and the modern trend towards “predictive policing”. This is the way a growing number of police forces around the world are using data on past crimes to predict the likelihood of crimes in the future. And the results have been generally favourable, with several forces claiming that it allows them to allocate resources more effectively. For example, if there have been burglaries in your neighbourhood in the recent past, the algorithm would flag it as a potential future crime scene. And that allows law enforcement organisations to react accordingly, such as patrolling high risk areas rather than low risk ones. The idea is to stop crime before it starts. But that raises an interesting question. How good can predictive policing become? And what kind of data will police forces use to make these improvements?
