Several cities are considering, testing, or already using statistical crime prediction software. Should we fear Tom Cruise and the Precogs? (More)
NPR reports that the Los Angeles police are testing statistical crime prediction software, similar to the system already in use in Memphis. Unlike the Phillip K. Dick story and Stephen Spielberg film The Minority Report, these do not rely on specially-trained people in vats applying psychic powers of precognition. Instead they use statistical models to predict where certain kinds of crimes are most likely to happen, so the police can put more resources in those areas:
UCLA anthropologist Jeff Brantingham says he’s not surprised [the software is effective]. Human behavior, especially when in search of resources, follows very predictable patterns. For his doctoral work, Brantingham studied foraging strategies of ancient hunter-gatherers in Mongolia.
“It’s surprising how similar the problems are,” he said. “How it is that ancient hunter-gatherers found gazelles on the Mongolian steppes is very similar to how it is that offenders find a car to steal.”
A human population has a minimum subsistence level for food. If a hunter-gatherer tribe can’t find enough food in one place, they’ll look somewhere else. Thus, Brantingham says, most crimes happen in predictable waves:
“If your house is broken into, then the chance that it is going to be broken into again goes way up, and the chance that your neighbor’s house is going to be broken into goes way up,” he said.
That’s because crooks now know the area and go back to where they had success. Brantingham says these crime waves show up in patterns similar to the aftershocks of an earthquake.
But the software may be even more effective than that. As Cass Sustein and Richard Thaler noted in Nudge, most crimes are opportunistic. If the characteristics of an area make it seem easy to steal a car or burgle a home without getting caught, more people in that area will steal cars or burgle homes. However, research shows that unlike a tribe seeking food, there is no minimum subsistence level for car thefts or home burglaries. If an area is made less inviting for a certain kind of crime, most opportunists in that area won’t go elsewhere. They stop committing that kind of crime.
Still, some civil libertarians worry about such technology:
While the science is impressive, Loyola Law School professor Stan Goldman says he worries how the data will be used. Despite police assurances, he says, authorities could use it as reasonable doubt [sic] to stop and search innocent suspects who happen to be in the highlighted neighborhoods.
“It may very well end up reducing crime to a certain degree,” he said. “The question is at what cost, at what price?”
(Note: The NPR reporter probably meant “reasonable suspicion” rather than “reasonable doubt.”)
Do you think police departments should use such tools to allocate resources or – better yet – to encourage local homeowners, businesses, and developers to make the area less inviting for crime?