Making the city safe for lawn chairs
The SPD’s new predictive policing software might predict crimes against garden gnomes, but it can’t prevent cops from using excessive force
Before officers in the Seattle Police Department (SPD) head out on a shift, they glance at a map on a computer at the precinct office.
The map is dotted with little red boxes. Each box represents an area where, according to an algorithm, crime is most likely to occur. Officers know to pay extra attention to those areas.
The software, called PredPol for predictive policing, is designed to forecast crimes much like meteorologists predict the weather. It’s loaded with crime reports stripped of everything except the time, date, location and type of crime. An algorithm crunches the data and selects several spots on a map where property crime is most likely to occur.
Data-driven policing is part of Mayor Mike McGinn’s SPD 20/20, a collection of 20 reforms to take place over the course of 20 months in response to a 2011 Department of Justice report that showed the SPD had a pattern and practice of excessive force.
This scientific analysis relies on computer-generated information to guide policing, which in turn is expected to build trust in the community by better responding to people’s needs.
But PredPol, which costs $45,000 a year, will be used to predict and prevent property crimes, a category of crime that has been on the decline almost every year for the last 25 years, according to the SPD. PredPol doesn’t address complaints raised by the DOJ report about a police culture that allowed officers’ use of excessive force to go unchecked.
Given these circumstances, how can predictive policing help reform the police?
SPD officials say it will build trust, which is a place to start. But City Councilmember Tim Burgess is circumspect.
“None of this stuff works unless there is a culture change,” Burgess, a former police officer, said of predictive policing. “And none of these things stand alone by themselves.”
Algorithms are the new black
Businesses increasingly rely on algorithms to crunch massive amounts of data in order to make decisions.
The Oakland Athletics famously used computer algorithms in 2002 to hire the best players on a limited budget. Companies such as Target use consumer data to determine what kind of advertisements to mail out to customers.
The Los Angeles Police Department (LAPD) teamed with University of California, Los Angeles mathematician George Mohler and anthropologist Jeff Brantingham to do similar number crunching for everyday police work.
LAPD officers saw results soon after using the tool in 2011. Property crimes dropped 12 percent in the Foothill area of Los Angeles.
“People are saying, how low can you go?” said LAPD Captain Sean Malinowski.
Time Magazine listed PredPol as one of its top 50 inventions in 2011.
Policing the predictors
Critics of predictive policing say its potential gains come at the risks of profiling poor communities and communities of color, reaffirming biases some say are already in effect in the SPD.
Police have always tried to select hot spots where they believe crime is prevalent. But data is not bias free, and past efforts to target high-crime areas have led to profiling, said Kade Crockford, director of Technology for Liberty Project.
If crime reports are biased in the first place, she said, removing race does not fix the problem.
“Data is not neutral; algorithms are not neutral,” Crockford said.
As SPD currently uses it, predictive policing is based on reported crime, said Lt. Bryan Grenon, who manages the program in Seattle. He acknowledged that some communities are more likely to report a crime than others. People who speak English as a second language are far less likely to turn to the police for help, for example.
“That’ll be an ongoing challenge,” Grenon said.
When McGinn announced the program May 17, he encouraged residents to call in any property crime, no matter how small. Report a theft, he said, even if it’s a missing lawn chair.
In Seattle, police will spend more time in the areas where lawn chairs are reported stolen, but that won’t necessarily mean more arrests, Grenon said.
As a result, it will be hard to tell how well predictive policing is working.
“A success is that no crime occurs,” Grenon said.
CommentsI am all for crime prevention of property theft, as long as it's not taking officers from more serious crime prevention. In that vein, I'd like to see about 2/3 less traffic cops (is there EVER one by when someone's on your bumper for a mile or passes you at breakneck speed while giving you the finger) and put on more serious crime prevention duties, but that's another subject. In respect to bias, that word is not the F bomb, we're all biased in one way or another, it's human nature. There's more crime in White Center than on Mercer Island, that's not bias, it's a fact. Crime ridden areas should have more police presence and if this software is a useful tool, then that's great.
Commenting is not available in this channel entry.