èƵ

Amazon face recognition mistakes US politicians for crime suspects

A face recognition service sold by Amazon falsely identified 28 members of Congress - and disproportionately black legislators - as people arrested for a crime
Two statues at US congress
Face recognition software matched US politicians to police mugshots
Kristoffer Tripplaar / Alamy Stock Photo

Face recognition technology sold by Amazon incorrectly matched 28 politicians with people arrested for a crime, an investigation by the American Civil Liberties Union has found. The findings raise further concerns about the use of similar technology by police departments in the US and beyond.

Amazon’s Rekognition is an image analysis service that offers the ability to automatically comb huge quantities of still pictures and video. It can detect objects and sensitive content, extract on-screen text, and compare faces to a reference image.

To perform their analysis, the ACLU built a database of 25,000 publicly available mugshots from police departments across the country. Researchers then programmed the Rekognition service to search that database for images which matched the official portrait of any sitting member of Congress, using the default settings offered by Amazon. According to the ACLU, commissioning the analysis cost $12.33 – “less than the price of a large pizza”.

The 28 false matches returned included politicians of all ages, ethnicities and from across the political spectrum. However, black legislators were disproportionately likely to be falsely matched with arrest photos. Six members of the Congressional Black Caucus were incorrectly matched, including civil rights leader John Lewis. Black legislators made up 40 per cent of the false matches, despite them making up only 20 per cent of Congress.

Technological bias

This is not the first time technology has been accused of racial bias. A ProPublica investigation found that an algorithm used in some US courts to help predict if someone will re-offend is more to over estimate the risk of black defendants re-offending, and the reverse for white defendants. Though this is disputed by the company who makes the software.

The use of face recognition technology is growing across the world, for anti-terrorism, market analytics, building access, and policing. However, critics warn that wide scale surveillance combined with a low accuracy in matches is threatening to undermine civil liberties.

Amazon is marketing Rekognition software to police forces in the US. Departments in Florida and Oregon are using the service to scour real time footage from CCTV cameras for faces matching mugshots on file.  The ACLU argues this a chilling effect on those exercising the right to assemble and protest, and puts innocent lives at risk.

“An identification — whether accurate or not — could cost people their freedom or even their lives,” the ACLU wrote in a statement. “People of color are already disproportionately harmed by police practices, and it’s easy to see how Rekognition could exacerbate that.”

The ACLU is calling for Congress to join them in demanding a moratorium on the use of face surveillance technology by police departments until more stringent safeguards can be put in place.

Amazon says that the ACLU analysis could have been improved by requiring higher confidence levels for a match. The ACLU used the default setting of 80 per cent. ” When using facial recognition for law enforcement activities, we guide customers to set a threshold of at least 95 per cent or higher,” wrote Amazon in a statement sent to èƵ.

“Finally, it is worth noting that in real world scenarios, Amazon Rekognition is almost exclusively used to help narrow the field and allow humans to expeditiously review and consider options using their judgement (and not to make fully autonomous decisions), where it can help find lost children, restrict human trafficking, or prevent crimes,” said Amazon.

Topics: Artificial intelligence / Privacy / Technology