Bach Seat

Back in January 2020, the Detroit Police Department arrested Robert Williams in his driveway in Farmington Hills according to The New York Times. He had his mug shot, fingerprints and DNA taken and was held overnight. Based on facial recognition software DPD decided that in October 2018 decided he had shoplifted 5 watches worth $3,800, from Shinola. Shinola is an upscale boutique that sells watches, bicycles, and leather goods in the trendy Midtown neighborhood of Detroit.

Mr. Williams knew that he had not committed the crime in question. What he could not have known, as he sat under arrest, is that his case may be the first known account of an American being wrongfully arrested based on a flawed match from a facial recognition algorithm, according to experts on technology and the law. This is part of the systemic racial bias in law enforcement that millions are protesting. They are protesting not just the actions of individual officers, but bias in the systems used to monitor communities and identify people for prosecution.

Facial recognition systems have been used by police forces for more than two decades. Recent studies by MIT. and NIST (PDF), have found that while facial recognition technology works relatively well on white men, the results are less accurate for other demographics, in part because of a lack of diversity in the images used to develop the underlying databases.

As part of this debate, IBM, Amazon, and Microsoft paused new sales of facial recognition systems to  law enforcement. The gestures were largely symbolic, given that the companies are not big players in the industry. The technology police departments use, according to the NYT, is supplied by companies that aren’t household names, such as Vigilant Solutions, Cognitec, NEC, Rank One Computing, and Clearview AI.

Clare Garvie, a lawyer at Georgetown University’s Center on Privacy and Technology, has written about problems with the government’s use of facial recognition told the NYT she suspects Mr. Williams’ case is not the first case to misidentify someone to arrest them for a crime they didn’t commit. “This is just the first time we know about it.”

Mr. Williams’ case combines flawed technology with poor police work, illustrating how facial recognition can go awry according to the New York Times. The original still unsolved Shinola shoplifting case occurred in October 2018. Katherine Johnston, a loss prevention contractor for Shinola reviewed the store’s surveillance video and sent a copy to the Detroit police, according to the DPD report. Where it sat until the Michigan State Police got involved – in a shoplifting case.

In March 2019, Jennifer Coulson, a digital image examiner for the Michigan State Police, uploaded a “probe image” — a still from the Shinola video, showing a man in a red Cardinals cap — to the state’s facial recognition database. The DataWorks Plus system mapped the man’s face and searched for similar ones in a collection of 49 million photos.

Since 2005 Michigan’s facial recognition technology has been supplied by a South Carolina company called DataWorks Plus under a contract worth $5.5 million. The NYT says DataWorks Plus does not formally measure the systems’ accuracy or bias. Todd Pastorini, a DataWorks Plus general manager told the NYT, “We’ve become a pseudo-expert in the technology.”

In Michigan, the DataWorks facial recognition software used by the state police incorporates components developed by the Japanese tech giant NEC and by Rank One Computing, based in Colorado, according to Mr. Pastorini and a state police spokeswoman. In 2019, algorithms from both companies were included in a federal study of over 100 facial recognition systems that found they were biased, falsely identifying African-American and Asian faces 10 times to 100 times more than Caucasian faces.

After MSP’s Coulson, ran her search of the probe image, the system would have provided a row of results generated by NEC and a row from Rank One, along with confidence scores. Mr. Williams’s driver’s license photo was among the matches. Ms. Coulson sent it to the Detroit police as an “Investigative Lead Report.” 

This is what technology providers and law enforcement always emphasize when defending facial recognition, says the article:  It is only supposed to be a clue in the case, not a smoking gun. DPD Chief James Craig describes himself as a “strong believer”  in facial recognition software.

Before arresting Mr. Williams, investigators could have sought other evidence that he committed the theft, such as eyewitness testimony, location data from his phone, or proof that he owned the clothing that the suspect was wearing. In this case, however, according to the Detroit police report, investigators simply included Mr. Williams’s picture in a “6-pack photo lineup” they created and showed it to Shinola’s loss-prevention contractor, and she identified him. Shinola’s contractor. Johnston declined to comment.

Rank One’s chief executive, Brendan Klare, found fault with Ms. Johnston’s role in the process. “I am not sure if this qualifies them as an eyewitness, or gives their experience any more weight than other persons who may have viewed that same video after the fact.”  John Wise, a spokesman for NEC, told the author: A match using facial recognition alone is not a means for positive identification.”

In Mr. Williams’s recollection, after he held the surveillance video still next to his face, the two detectives leaned back in their chairs and looked at one another. One detective, seeming chagrined, said to his partner: “I guess the computer got it wrong.” They turned over a third piece of paper, which was another photo of the man from the Shinola store next to Mr. Williams’s driver’s license. Mr. Williams again pointed out that they were not the same person.

Mr. Williams asked if he was free to go. “Unfortunately not,” one detective said. Mr. Williams was kept in custody for 30 hours, and released on a $1,000 personal bond. The Williams family contacted defense attorneys, most of whom, they said, assumed Mr. Williams was guilty of the crime and quoted prices of around $7,000 to represent him. They, also tweeted at the American Civil Liberties Union of Michigan, which took an immediate interest. said Phil Mayor, an attorney with the organization told the NYT:

We’ve been active in trying to sound the alarm bells around facial recognition, both as a threat to privacy when it works and a racist threat to everyone when it doesn’t,”  “We know these stories are out there, but they’re hard to hear about because people don’t usually realize they’ve been the victim of a bad facial recognition search.

Two weeks later, Mr. Williams appeared in a Wayne County court for an arraignment. When the case was called, the prosecutor moved to dismiss, but “without prejudice,” meaning Mr. Williams could later be charged again. Maria Miller, a spokeswoman for the prosecutor, said a second witness had been at the store in 2018 when the shoplifting occurred but had not been asked to look at a photo lineup. If the individual makes an identification in the future, she said, the office will decide whether to issue charges.

A DPD spokeswoman, Nicole Kirkwood, said that for now, the department “accepted the prosecutor’s decision to dismiss the case.” In a second statement to the NYT DPD doubled down saying it, “does not make arrests based solely on facial recognition. The investigator reviewed the video, interviewed witnesses, conducted a photo lineup.”

The ACLU of Michigan filed a complaint with the city (PDF),  asking for an absolute dismissal of the case, an apology, and the removal of Mr. Williams’s information from Detroit’s criminal databases.

Mr. Williams’s lawyer, Victoria Burton-Harris, said that her client is “lucky,” despite what he went through. Ms. Burton-Harris said to the NYT, 

He is alive … He is a very large man. My experience has been, as a defense attorney, when officers interact with very large men, very large black men, they immediately act out of fear. They don’t know how to de-escalate a situation.

Mr. Williams had an alibi, had the Detroit police checked for one.

rb-

Just to celebrate Independence day – the Georgetown Law’s Center on Privacy and Technology says, at least a quarter of the nation’s law enforcement agencies have access to face recognition tools. The MSP database has almost 50 million pictures in it for about 8 million adults in Michigan. That is over 6 pictures per adult Michigander – many come from the Secretary of State when you get a driver’s license but undoubtedly many are scrapped from social media sites. Michigan is one of at least 16 states that allow the FBI to search its database of driver’s license photos.

While the MSP didn’t start using facial recognition technology until 2001, the Secretary of State’s Office has been giving State Police all its digital photos — without notice to motorists — since 1998.

DataWorks provides facial recognition systems to both DPD and MSP. The DPD two-year $1 million contract for the DataWorks Plus software is set to expire in July 2020. Detroit City Council President Brenda Jones told the Detroit News that the police department agreed to pull back its most recent request for a contract extension and conduct community outreach before seeking approval to extend the contract through Sept. 30, 2022.

Dan Korobkin, deputy legal director for the ACLU of Michigan points out that Civil rights icon Martin Luther King Jr. “was the target of massive FBI surveillance, under what was then the latest state-of-the-art technology.” In response, Robert Stevenson, executive director of the Michigan Association of Chiefs of Police and retired chief of the Livonia Police Department, told GovTech he believes most Michiganders trust the police, “We’ve evolved in the last 50 years, as a country, and as police agencies.” Well just ask George Floyd.

Stay safe out there!

Related article

• New screen inspired by butterflies reduces eye strain (Electronics 360)

Ralph Bach has been in IT long enough to know better and has blogged from his Bach Seat about IT, careers, and anything else that catches his attention since 2005. You can follow him on LinkedIn, Facebook, and Twitter. Email the Bach Seat here.