Emily Badger is a former staff writer at CityLab. Her work has previously appeared in Pacific Standard, GOOD, The Christian Science Monitor, and The New York Times. She lives in the Washington, D.C. area.
Could better software change how police departments find fugitives?
The job of a warrant officer sounds pretty maddening. On any given day in a major metro area, a single officer may be working against hundreds of offenders, all accumulated on a list that puts homicide fugitives alongside traffic scofflaws. Then you've got all those people actively trying not be found. Go to some guy's address, and his mom will tell you he's working at Burger King. Go to Burger King, and he's apparently "on vacation." Or better yet, find the guy, and half the time he'll tell you, "nope, that's my brother!"
"That happens a lot," says Nick Selby (this is why you've got to pull a photo of the guy first). Selby and David Henderson, both sworn officers in the Dallas-Fort Worth area, have been trying to design a software solution to the warrant officer's quandary. Their start-up, called StreetCred Software, is one of five companies recently selected for Code for America's newest accelerator class, a group of entrepreneurs trying to leverage government data of various kinds in in the expanding "civic technology" sector.
Some of the participating start-up ideas aim to help streamline government procurement decisions, register new businesses, comply with public-records requests, and aid family-support services. Selby and Henderson want to make it easier to find fugitives. The concept came out of Henderson's own experience as a warrant officer.
"He spent a lot of time doing really stupid things, making two print-outs from a dozen different websites for each fugitive he went after," Selby says. "And he’s going after thousands, organizing these things by folders put into milk crates labeled by zip code."
This situation sounds like classic government IT, where a public servant seemingly has less sophisticated technology than the average teenager. "In our experience," Selby says, "cops have green screens, no maps, no images, all text. And it’s one database at a time."
Henderson wanted to put all of these people on a map, which turned out to be the easy part. The harder part was pulling in all the information that an officer might normally spend a few hours researching and photocopying at a desk: driver's license photos, court records, vehicle and property details, and more. Some police software is already doing this kind of work. But StreetCred's innovation, specifically targeted at warrant officers, is that it can use all that data to calculate probabilities about how likely an officer is to actually find the person attached to an outstanding warrant at the address officials have for him.
"Instead of giving an officer a list of 'here’s a thousand fugitives you have to catch, go get them,' it’s 'hey, in the next five hours, you can probably get these five people,'" Selby says. "That’s a much more manageable job."
Some of the computational work seems pretty obvious. The system wants to highlight the "good" fugitives and throw away the "bad" ones, meaning those people who are highly unlikely to be caught for ridiculous reasons – they're dead, or were jailed last week (again, these are both common scenarios). But from there, the algorithms grow more nuanced, calculating probabilities based on the type of crime, the type of offender, what car the person drives or job they're known to have.
Because all of these people have outstanding warrants against them – a judge has already said, "bring him in" – the software can operate from a set of assumptions and calculations that aren't available to police who might otherwise have to articulate reasonable suspicion or probable cause. But the software doesn't, Selby says, consider race or gender.
Henderson and Selby both began working on the company full-time about six months ago, although they still work part-time as police officers. They began selling StreetCred to some Texas and New York state police departments in November of last year (Selby counts "fewer than 25" agencies). Selby is quick to stress that the tool doesn't "do police work," rather it does the heavy clerical lift that makes police work possible. That may be an important distinction for anyone (old-school officers included) who's uncomfortable with the idea of algorithms involved in catching criminals.
So far, Selby says the officers who've seen the tool quickly recognize its value.
"People who do government work and have to use government computers, they might not be able to tell you what they need or why they’re not able to do what they want to do," he says. "But when they see something that does it, they know it immediately."
Top screengrab courtesy StreetCred.