How facial recognition has become a routine police tool in America

But these systems proliferate as facial recognition is growing in concern – researchers in artificial intelligence and privacy have discovered that algorithms at the root of some systems incorrectly identify women and people with dark skin more frequently than white men – and allow the government to extend surveillance. from the public without much supervision. Some agencies have policies on the use of facial recognition, but there are few laws or regulations governing the databases that systems can draw from, the people in those databases, the circumstances under which the police can scan people's photos, the accuracy of systems, and how much the government should share with the public about its use of technology.

Police praise the power of technology to improve investigations, but many agencies are also trying to keep their methods secret. In New York, the police department has resisted attempts by defense lawyers and privacy advocates to reveal the workings of its facial recognition system. In Jacksonville, Florida, the authorities refused to disclose the details of their research in facial recognition to a man who defended his conviction for selling $ 50 worth of crack. Sometimes people arrested with the help of facial recognition do not know that it has been used against them.

Since the police do not consider facial recognition as evidence to be presented in court, the technique does not often appear in public documents and has not been the subject of many court decisions. Its use and distribution are difficult to follow.

The companies that build the technology are also struggling with the consequences of its use. Amazon has made its facial recognition system available to the police, prompting protests from employees, shareholders and researchers in artificial intelligence. Microsoft claims to have resisted requests to sell its products to the police and asked for government regulation. Axon, the largest body camera manufacturer in the United States, has filed patents for facial recognition applications, but says it does not pursue them because it consults an ethics committee on artificial intelligence.

At the same time, companies are creating even more advanced systems that will allow police to identify people from live video footage, such as body cameras, rather than just still images. It's only a matter of time before such technology is available to the police.

These developments have triggered attempts to limit the use of facial recognition by the police. The cities of Somerville, Massachusetts, and San Francisco and Oakland, California, plan to ban it. So is the state of Massachusetts.

Civil rights advocates, privacy researchers and criminal defense lawyers warn that the ubiquity of police-managed facial recognition systems could incite officers to become overly dependent on a criminal justice system. defective technology and risking wrongful convictions. According to them, this could trigger an explosion of arrests for minor offenses, exacerbate the disproportionate impact of the criminal justice system on the poor and minorities, and lead to even more common uses already deployed in China such as as the exhibition of geawalkers and people taking toilet paper public toilets.

"This could have a panoptic effect if you are concerned that the government is constantly monitoring," said Jake Laperruque, senior advisor to Project on Government Oversight, a non-profit organization that investigates the federal government.

"We try to use it as much as possible"

As these debates unfold, police continue to adopt facial recognition technology, a trend that began two decades ago when Pinellas County, Florida won a series of federal grants to become test ground for emerging technology. Sheriff Bob Gualtieri said the technology has changed the police almost entirely for the better, allowing his investigators to identify bank robbers, missing persons, even those killed in car crashes. "We are solving crimes that we would not have solved otherwise," Gualtieri said.

The technology has since crossed the country, including Los Angeles, San Diego, Chicago and New York, as well as hundreds of national and local law enforcement agencies.

Since there is no easy way to measure the number of police services that have adopted facial recognition, any effort to do so provides only a snapshot. The most comprehensive evaluation was conducted by the Georgetown Center for Privacy and Technology, which found in 2016 that at least one in four police departments could conduct facial recognition research, either through through a system they bought themselves, from a system owned by another organization. (For example, the Pinellas County system, which includes millions of driver's licenses and photos of Florida law enforcement, is available to nearly 300 other law enforcement agencies.) Technology has undoubtedly spread since then.

In Colorado, local investigators have foiled credit card fraudsters, bandits on power tools and garage burglars at home, and identified suspects in a shooting and road rage case. In San Diego, police take pictures of suspicious people in the field who refuse to identify. This technology captured a serial thief in Indiana, a Pennsylvania rapist, a car thief in Maine, robbery suspects in South Carolina, a sock thief in New York, and thieves in New York. stranger in Washington County, Oregon.

In southwestern Ohio, police are throwing images of Crime Stoppers alerts into their newly acquired facial recognition system and solving all kinds of property crimes.

"We try to use it as much as possible," said Jim Stroud, a police inspector from Springfield Township, in the suburbs of Cincinnati.

In Lakewood, Colorado, Det. Mark Gaffney patrols with a mobile phone with an app that allows him, during roadside checks and roadside encounters, to take pictures of people whose names he can not confirm and pass them through. facial recognition system. Almost instantaneously, the program provides a list of potential matches containing information that may help to confirm the identity of those arrested and to determine if there are outstanding warrants. Previously, he had to let the person go or bring him in with his fingerprints.

"When I do not know who a person is, I have more options available to me if she does not want to tell me," Gaffney said.

Inside a department

Few local law enforcement agencies openly explain how they use facial recognition. Exceptions include the Arapahoe County Sheriff's Office, which allowed investigators to describe the integration of technology into their daily files.

The application of facial recognition by the agency is typical among law enforcement agencies, with a specially trained group of investigators acting as liaisons with the rest of the department. Every day, they sit in front of their computer at the agency's headquarters in Centennial, Colorado, to insert images into their one-year-old face recognition system. The photos come from their own records, Crime Stoppers newsletters and big box retailers such as Home Depot and Target, hit by serial thieves. They also receive requests from detectives from other police departments.

The amount of material is huge, especially for property crimes, because, generally speaking, if there is a place that is worth taking – a street, a shop, a house – it There is probably also a surveillance camera or a witness.

The images are downloaded into software called Lumen, sold by Numerica Corp., a subcontractor specializing in defense and law enforcement. An algorithm compares them to a database of photo IDs and photo reservations shared by law enforcement agencies in the Denver area.

In a few seconds, dozens of possible matches are displayed on the screen, depending on their similarity, according to the algorithm. Investigators scour the list, looking for a potential match, adding filters such as sex, race and eye color and subject hair. The confidence of the algorithm in the match never reaches 100%. If the subject's photo is grainy or has a face on the side, or a hood or sunglasses, the results may not be of interest. Sometimes investigators find promising success at the top of the rankings; at other times, the suspect appears far behind, with perhaps 75%.

Any apparent connection must then be verified. The investigator goes to work and performs the potential correspondence in criminal databases for more clues. The interviewer can also set up an array of pictures, including potential correspondence and pictures of people of similar appearance, to present to a witness.

"Facial recognition recognition is just a track of investigation. This does not establish the probable cause of arrest, "said Rick Sheets, an Arapahoe County investigator specializing in property crime.

Lumen-trained Arapahoe investigators encourage their colleagues not to put the files away before trying. This is what happened in March 2018, when investigator Jim Hills spoke to Young about the system, which had been installed a few weeks earlier.

Young thought about the woman who was robbed $ 400 at the bowling appointment seven months ago.

"It could be a good one to try," she said.

A mystery solved

Young gave Hills a close-up photo of the date thief on his cell phone. The victim, Debbie Mallick, had communicated it at the beginning of the investigation, as well as screenshots of text messages in which he confessed to having stolen $ 400, just before stopping his phone.

Hills went into the snapshot in Lumen. Near the top of the results was an old picture of a man with dozens of previous arrests, with convictions for assault, car theft and violation of orders not to do. He looked remarkably similar. His name was Antonio, the same as the thief's.

Young is arranged for a colleague to show Mallick a picture board. Mallick chooses him immediately. Young has filed a warrant arresting the man, Antonio Blackbear, for theft for misdemeanor.

It was two months before Young received an alert from the Denver police, who said he had taken Blackbear into custody. In June 2018, he pleaded guilty, was sentenced to 25 days in prison and reimbursed Mallick.

Blackbear, 40, could not be reached for comment.

Mallick, 47, said she still had not recovered her money, which was all she wanted in the first place. The experience has deeply hurt her and made her less confident towards people. But the mother of two said she was grateful to the investigators for not giving up.

She did not know how the police had identified Blackbear.

In the arrest reports and court documents detailing the case, there is no mention of facial recognition.

Jon Schuppe

Jon Schuppe writes about crime, justice and related topics for NBC News.