Why aren’t California lawmakers banning police from using facial recognition technology? | Opinion

Nate Freed Wessler
·4 min read

To date, we know of seven wrongful arrests in the United States caused by incorrect face recognition results — and those are only the cases that have become public. In six out of the seven cases, the wrongfully accused individuals were Black, a figure that corresponds to numerous studies showing that facial recognition technology misidentifies Black people and other people of color at higher rates than white people. This is due, in part, to biases in the photo databases used to train the algorithms.

Despite these failures, however, police departments across the state and the country are increasingly using facial recognition technology to try to match pictures of suspects with driver’s license photos, mugshots and images from other databases.

Opinion

A technology this biased and error-prone should not be used by the police. Unfortunately, California has responded with Assembly Bill 1814, authored by Phil Ting, D-San Francisco. The bill is a misguided attempt to regulate — rather than prohibit — police use of facial recognition by merely declaring that officers shouldn’t rely on this technology as the sole basis to obtain an arrest warrant. While this might sound sensible, it will not stop wrongful arrests. In fact, bills like AB 1814 will make the problem worse.

Even a short time in jail can turn a person’s life upside down. Wrongful arrests caused by facial recognition have set people back thousands of dollars, costing them their job and their home and inflicting lasting psychological and emotional harm on their families.

Last year, for example, six Detroit police officers showed up at the doorstep of Porcha Woodruff, who was eight months pregnant. She was arrested as she was preparing her children for school. Woodruff ended up in the hospital after being released, receiving two bags of intravenous fluid for dehydration. The charges against her were later dropped, and police admitted they had erred, but the damage was done.

At the American Civil Liberties Union, we’ve warned that facial recognition technology is not going to stop producing false matches, no matter the standards lawmakers may set or the updates companies make to their algorithms. There are simply too many variables — from the photo database, to the settings, to the picture quality — which can allow things to go wrong.

Police from Los Angeles County, to Detroit, to the U.S. Department of Homeland Security have tried to resolve this issue by requiring officers to independently confirm the results to establish probable cause. This is the same approach taken by AB 1814.

But there’s a problem: in most of the known cases of face recognition wrongful arrests — including in Woodruff’s case — police did try to confirm the match, but then arrested the wrong person anyway.

Too often, when police take additional investigative steps, those steps exacerbate and compound the unreliability of facial recognition technology. Rather than being an asset to police investigations, facial recognition poisons them.

This technology generates a list of faces that are similar to the suspect’s image, but frequently will not actually be a match. When police try to confirm the match (often through a photo lineup), it presents the witness with an image of a doppelgänger. It should come as no surprise that if you present a photo nearly identical to the suspect, the witness is likely to choose that image.

This is exactly what happened with Woodruff and others, including Harvey Eugene Murphy Jr., who recently filed a lawsuit alleging that he was wrongfully arrested by Houston police last year based on a photo lineup tainted by an incorrect face recognition result.

Instead of preventing false matches, half measures like AB 1814 codify the very process that produced them. The last thing we need is for the legislature to tell the police that if they follow the steps in this bill, the technology is now safe to use.

There is no safe way for police to use facial recognition. It must be strictly curtailed — and the best way to do this is through strong bans. More than 20 jurisdictions across the country, from San Francisco to Pittsburgh and Boston, have done just that.

If AB 1814 becomes the legislature’s answer to the false-match problem, more wrongful arrests are all but guaranteed.

Nate Freed Wessler is the deputy director of the ACLU National’s Speech, Privacy and Technology Project.