Predictive Modeling as Crime Prevention
Most police departments provide datasets to the public, but until recently, few have managed to do anything particularly useful with it.
Predictive policing–the ability to identify potential criminal activity through predictive modeling–could be the greatest weapon in a police department’s arsenal.
Vancouver, Canada has already deployed predictive crime mapping tools to police units.
Modesto police officials credit predictive policing with a double-digit drop in burglaries, robberies and vehicle thefts in the past 10 months, and they credit the department’s new policing strategies for the city’s lowest property crime rates in three years.
In a move meant to bring technological prowess to the Seattle Police Department, SPD recently named Amazon vice president Greg Russell as its new chief information officer.
Beyond overseeing the SPD’s internal technology systems, Russell is tasked with finding new ways of using data to track and predict crime, while balancing those capabilities with the privacy concerns of citizens. The department has been grappling with issues related to police body cameras, video disclosure, drones and other forms of emerging technology.
In an SPD Blotter interview, Russell foreshadows his approach in the CIO’s role. “Transparency to me means you’re being brutally honest with the information,” he said. “You’re not trying to sway it one way or the other. The data is the data. You’re just making it available.”
Russell: “You have to understand what information you can’t share,” Russell said. “But beyond that, the more transparent you can be with information, the better. If you can show the public that you’re sharing the good news and the bad news, you become more trustworthy … My understanding is that we have a lot of data, we just can’t really make it available. It’s not that we don’t want to. It’s just that we don’t know how to.”
Jay Feng, a University of Washington engineering major, found a way to do exactly that–he made the data available in a way that facilitates action. As featured in GeekWire, Feng used data to build a crime-predicting model for Seattle.
Using Seattle government data open to the public, with the help of Socrata’s open data solutions, Feng says the most popular page is “Seattle Police Department 911 Incident Response,” which at the time of this report, had more than 50,000 views.
Another addictive and useful real-time rabbit hole, My Neighborhood Map, relies on open data to indicate recent crimes in your neighborhood.
Several predictive policing methods are currently in use in law enforcement agencies across the United States, and there are some signs that the strategy may be working; as Forbes notes, early results in cities like Atlanta, Los Angeles and Santa Cruz saw both reduced crime rates and better prediction compared to conventional analysts.
Does positioning police in hotspots discourage opportunistic wrongdoing while encouraging other criminals to move to less likely areas? Can predictive modeling move police into proactive rather than reactive roles? It will be interesting to see how Greg Russell reshapes the SPD’s arsenal.