Detroit police chief cops to 96-percent facial recognition error rate

Detroit’s police chief admitted on Monday that facial recognition technology used by the department misidentifies suspects about 96 percent of the time. It’s an eye-opening admission given that the Detroit Police Department is facing criticism for arresting a man based on a bogus match from facial recognition software.

Last week, the ACLU filed a complaint with the Detroit Police Department on behalf of Robert Williams, a Black man who was wrongfully arrested for stealing five watches worth $3,800 from a luxury retail store. Investigators first identified Williams by doing a facial recognition search with software from a company called DataWorks Plus. Under police questioning, Williams pointed out that the grainy surveillance footage obtained by police didn’t actually look like him. The police lacked other evidence tying Williams to the crime, so they begrudgingly let him go.

Now Vice’s Jason Koebler reports that Detroit Police Chief James Craig acknowledged the flaws with its facial recognition software at a Monday event.

“If we would use the software only [to identify subjects], we would not solve the case 95-97 percent of the time,” Craig said. “That’s if we relied totally on the software, which would be against our current policy … If we were just to use the technology by itself, to identify someone, I would say 96 percent of the time it would misidentify.”

As Craig notes, police officers in Detroit aren’t supposed to arrest someone based solely on the results of a facial recognition search. And the Detroit police claimed that they didn’t do that in the Williams case. A police spokeswoman told The New York Times that “the investigator reviewed video, interviewed witnesses, [and] conducted a photo lineup” before arresting Williams.

But this “investigation” was rather flimsy. The “photo lineup” consisted of showing photos to a security contractor who was not an eyewitness; he had only viewed the same surveillance footage police had used in the first place.

Research has found that the accuracy of facial recognition software varies by the race of the subject, with Black suspects being identified less often than white ones. And Koebler points out that the DPD’s own statistics show the technology being used almost exclusively on Black suspects. According to police data, 68 out of 70 facial recognition searches were done on Black suspects, while two had a race code of “U”—probably short for “unknown.”

The ACLU has called on the Detroit Police Department—and other police departments—to stop using facial recognition technology for investigations in light of its high error rate and racially disparate impact. Boston’s city council voted to ban the use of facial recognition technology last week.

https://arstechnica.com/?p=1688306