Predicting where crime will take place
July 5, 2012 1 Comment
Cop shows up for roll call before patrolling the beat. On her department-issued tablet is a map showing where on her beat crime will likely occur. The map is annotated by notes from veteran cops.
She spends a little more time in those spots, maybe talks up some of the locals about what’s happening. Maybe the predictions heighten her observations of people in these areas and she learns something actionable. Maybe just by being there she cools tensions. Maybe some of her chats with the residents dissuades them from something. Maybe crime declines by chance.
That’s not some deleted scene from Minority Report, or a scenario from Person of Interest, that’s happening right now, at least according to the marketing hype of PredPol’s site. Its customers claim that it is better than human-based prediction, and a useful tool.
Has this been properly tested to minimize the possibility that the results are not a matter of random chance, or that the Hawthorne Effect is in play? The Predpol site has some results, but they don’t come close to satisfying a social scientist. If it does work, what happens when programs like it improve? And what if it can expand beyond property crimes? What if that cop visits a house the day before a domestic disturbance would occur and prevents an assault or even a death?