Heavily pregnant woman jailed for robbery for false facial recognition

Detroit
Pre-trial detention in the eighth month: heavily pregnant woman is booked for robbery because of false face recognition

Again and again, AI experts warn of the consequences of facial recognition software. A woman has now been arrested for the first time in the United States for false detection. And was in custody with labor pains.

Being heavily pregnant as a mother of two children is actually enough of a challenge. Due to a miscarriage of justice, the 32-year-old Porcha Woodruff had to experience even more suffering. Despite her pregnancy, she became one because of the allegation arrested for robbery and auto theft. Because a police AI thought she was the perpetrator.

When police showed up at her door last February, Woodruff initially thought she was being kidded, she tells The New York Times. “Are you kidding me,” she asked the police officers who were standing in front of her house in Detroit as she was about to take her children to school. She drew attention to her pregnancy belly, tried to explain how absurd the idea of ​​committing a robbery in the eighth month was. But the police couldn’t be persuaded. She was taken away in handcuffs in front of her children.

Eleven hours in custody

Only later did she learn the reason for her arrest. The Detroit Police Department had used facial recognition software that Woodruff had identified in a database as the perpetrator. Apparently that was enough for the court. After 11 hours in custody, Woodruff was allowed out of the building on $100,000 bail. Weeks later, the lawsuit was dropped due to lack of evidence.

Because of her pregnancy, the burden on the innocent accused was particularly high. “I went into labor from the stress in the cell. My back hurt, I had cramps,” she says. After being released, she decided to go to the hospital. There she was diagnosed with dehydration and was given fluids intravenously. Until the end of the pregnancy, she suffered extremely from the stress and the two interviews conducted by the police via video telephony.

Known AI issue

Woodruff isn’t the first person to be arrested in the United States for facial recognition errors. According to the NYT, this has already happened three times to the Detroit Police Department alone. However, Woodruff is the first woman in which such a misidentification is known. And: As in all other known cases of false AI detection, she is African American.

It has been a known problem for years that AI programs are particularly error-prone in black people. Since the AI ​​programs are trained with ready-made data sets and black people are severely underrepresented in the software industry, the programs seem to have trained a problem with darker skin tones when recognizing patterns in faces. A study by the US elite university Harvard had already clearly presented corresponding indications in 2020. Across the board, people with darker skin and women in particular were poorly recognized there.

This is a particularly serious problem in the context of the police: According to the authority’s transparency report, the majority of the 125 facial recognition searches retrieved annually by the Detroit Police Department related to black suspects. The police department “has every reason to know the risks of using facial recognition,” Clare Garvie, a technology expert at the Criminal Lawyers’ Association, told the newspaper. “And it still happens.”

lawsuit against the city

The Woodruff case is now occupying the courts elsewhere. The justice victim has filed a lawsuit against the city over her arrest. The allegations are indeed “very worrying,” the police chief said. You take the matter very seriously. The prosecutor would apparently do the same thing again. She continues to consider the charges “reasonable given the facts,” her office said.

Source: New York Times, Harvard study