Tech­nol­ogy is trans­form­ing the craft of polic­ing

The Denver Post - - TECH KNOW - By Joshua Brustein Bloomberg News

There’s a story Brett Gold­stein likes to tell. It starts on a Fri­day night in 2010 with him sit­ting in a dark­ened Crown Vic­to­ria on a Chicago street, por­ing over maps. Gold­stein was a com­man­der at the Chicago Po­lice Depart­ment, in charge of a small unit us­ing data analysis to pre­dict where cer­tain types of crimes were likely to oc­cur at any time.

Ear­lier that day, his com­puter mod­els forecast a height­ened prob­a­bil­ity of vi­o­lence on a par­tic­u­lar South Side block. Now that he and his part­ner were there, Gold­stein was doubt­ing him­self.

“It didn’t look like it should be a tar­get for a shoot­ing,” he re­called. “The houses looked great. Ev­ery­thing was well man­i­cured. You ex­pect, if you’re in this neigh­bor­hood, you’re look­ing for aban­doned build­ings, you’re look­ing for peo­ple sell­ing dope. I saw none of that.”

Still, they staked it out. Gold­stein’s wife had just given birth to their se­cond child, and he was ex­hausted af­ter a day in the of­fice. He started to doze off. Gold­stein’s part­ner ar­gued that the data must be wrong. At 11 p.m., they left.

Sev­eral hours later, Gold­stein woke up to the sound of his Black­Berry buzzing. There had been a shoot­ing — on the block where he’d been camped out.

“This sticks with me be­cause we thought we shouldn’t be there, but the com­puter thought we should be there,” Gold­stein said. He took it as vin­di­ca­tion of his vi­sion for the fu­ture of law en­force­ment. “I do be­lieve in a po­lice­man’s gut. But I also be­lieve in aug­ment­ing his or her gut.”

Seven years af­ter that evening, Gold­stein threw on a gray suit and headed from his Man­hat­tan ho­tel to New Jer­sey. Last spring he founded CivicS­cape, a tech­nol­ogy com­pany that sells crime-pre­dict­ing soft­ware to po­lice de­part­ments. Nine cities, in­clud­ing four of the coun­try’s 35 largest cities by pop­u­la­tion, are us­ing or im­ple­ment­ing the soft­ware, at an an­nual cost from $30,000 for cities with less than 100,000 peo­ple to $155,000 in cities with pop­u­la­tions over 1 mil­lion. Gold­stein was check­ing in on the two clients who were fur­thest along: the po­lice de­part­ments in Cam­den and Lin­den.

Gold­stein’s ped­dling some­thing that ev­ery depart­ment is af­ter nowa­days: tech­no­log­i­cal so­phis­ti­ca­tion.

Crit­ics in­side, out­side de­part­ments

Not ev­ery­one is rub­bing their hands in an­tic­i­pa­tion. Many po­lice of­fi­cers still see so-called pre­dic­tive polic­ing soft­ware as mumbo jumbo. Crit­ics out­side law en­force­ment say it’s ac­tively de­struc­tive. The his­tor­i­cal in­for­ma­tion these pro­grams use to pre­dict pat­terns of crime aren’t a neu­tral re­count­ing of ob­jec­tive fact; they’re a re­flec­tion of so­cioe­co­nomic dis­par­i­ties and the ag­gres­sive polic­ing of black neigh­bor­hoods. Com­puter sci­en­tists have held up pre­dic­tive polic­ing as a poster child of how au­to­mated de­ci­sion mak­ing can be mis­used. Oth­ers mock it as pseu­do­science.

“Sys­tems that man­u­fac­ture un­ex­plained ‘threat’ as­sess­ments have no valid place in con­sti­tu­tional polic­ing,” a coali­tion of civil rights and tech­nol­ogy as­so­ci­a­tions, in­clud­ing the ACLU, the Bren­nan Cen­ter for Jus­tice, and the Cen­ter for Democ­racy & Tech­nol­ogy, wrote in a state­ment last sum­mer.

The po­lice de­part­ments Gold­stein vis­ited in New Jer­sey didn’t raise any ques­tions about fair­ness dur­ing his re­cent trip — but there was skep­ti­cism nonethe­less. He had barely started speak­ing to a group of top of­fi­cers in the Lin­den Po­lice Depart­ment when the man who han­dled the city’s pro­cure­ment process con­fessed how wary he was of soft­ware ven­dors’ mag­i­cal-sound­ing claims. Gold­stein nod­ded. As a cop, he said, he hated sit­ting through “the ven­dor non­sense.” Gold­stein launched into a sing-song voice: “Oh, you’re go­ing to have a fly­ing car, and it’s go­ing to stop peo­ple, and you’re go­ing to be Su­per Po-Po!’ They’ll prom­ise you any­thing.”

Gold­stein’s com­pany does make one un­usual prom­ise, which it thinks can sat­isfy skep­tics in law en­force­ment and civil rights cir­cles si­mul­ta­ne­ously. Other com­pa­nies that make pre­dic­tive soft­ware for crim­i­nal jus­tice set­tings keep their al­go­rithms se­cret for com­pet­i­tive rea­sons. In March, CivicS­cape pub­lished its code on GitHub, a web­site where com­puter pro­gram­mers post and cri­tique one another’s work. The un­prece­dented move caused an im­me­di­ate stir among peo­ple who fol­low the cop tech in­dus­try.

“They’re do­ing all the things I’ve been scream­ing about for years,” said An­drew Fer­gu­son, a pro­fes­sor at the Univer­sity of the District of Columbia’s law school and author of the forth­com­ing book, “The Rise of Big Data Polic­ing.”

Post­ing com­puter code on­line won’t erase the wor­ries about pre­dic­tive polic­ing. There are still con­cerns about how CivicS­cape re­sponds to per­ceived short­com­ings, and there’s also the big ques­tion of what po­lice de­part­ments do with the in­tel­li­gence it pro­duces.

Pho­tos by Misha Fried­man for Bloomberg

Cam­den of­fi­cer Vi­dal Rivera on pa­trol in the neigh­bor­hood where he grew up.

An an­a­lyst ex­am­ines in­com­ing data, try­ing to pre­dict where of­fi­cers should be de­ployed, at Real Time Tac­ti­cal Op­er­a­tion In­tel­li­gence Cen­ter in Cam­den, New Jer­sey.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.