Government

Facial recognition software misidentifies Loebsack as criminal

ACLU ran test comparing faces with 25,000 arrest photos

Rep. Dave Loebsack, D-Iowa City, speaks at a June political event in Cedar Rapids. Loebsack was among 28 members of Congress that facial recognition software incorrectly identified as someone who had been arrested for a crime, according to the ACLU, which conducted the test. (Rebecca F. Miller/The Gazette)
Rep. Dave Loebsack, D-Iowa City, speaks at a June political event in Cedar Rapids. Loebsack was among 28 members of Congress that facial recognition software incorrectly identified as someone who had been arrested for a crime, according to the ACLU, which conducted the test. (Rebecca F. Miller/The Gazette)

Mark Twain once wrote “There is no distinctly American criminal class — except Congress,” but an experiment that identified members of Congress as criminals was a case of mistaken identity.

In a test of Amazon’s facial recognition software, Rekognition, conducted by the ACLU, Iowa 2nd District Democratic Rep. Dave Loebsack was among those incorrectly identified as someone who had been arrested for a crime.

The ACLU used the software Amazon markets to law enforcement agencies to compare mugshots of 535 members of Congress against 25,000 publicly available arrest photos.

Loebsack was not alone. Three senators and 25 House members were incorrectly matched as other people who have been arrested.

While some people may be upset about members of Congress being misidentified, Loebsack, an Iowa City Democrat, said his concern is that “this can happen to real people.”

The “real people” impact, according to the ACLU, is that a law enforcement officer could be biased toward a person before an encounter begins if facial recognition software misidentifies that person as someone with a criminal record.

A person could be questioned or have his or her home or vehicle searched based on false identification, the ACLU said.

ARTICLE CONTINUES BELOW ADVERTISEMENT

“With each new advance in technology that is used to identify people, we must be extremely careful in dealing with the privacy issues that surround it,” Loebsack said. “We have to figure out the correct balance to ensure no one is falsely accused, but also allow the police do their job and keep the public safe.”

Amazon pointed out that the ACLU set the confidence threshold for the test at 80 percent. The 80 percent threshold is appropriate for everyday objects, Amazon said, but it recommends a setting of 95 percent to identify individuals with a “reasonable level of certainty.”

The ACLU said the false matches were disproportionately of people of color, including six members of the Congressional Black Caucus.

In a recent letter to Amazon CEO Jeff Bezos, the black caucus expressed concern about the “profound negative unintended consequences” that face surveillance could have for black people, undocumented immigrants and protesters.

These results demonstrate why Congress should join the ACLU in calling for a moratorium on law enforcement using facial surveillance software, the ACLU said.

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

To see the 28 members of Congress who were incorrectly identified as criminals, go to aclu.org.

l Comments: (319) 398-8375; james.lynch@thegazette.com

CONTINUE READING

Give us feedback

We value your trust and work hard to provide fair, accurate coverage. If you have found an error or omission in our reporting, tell us here.

Or if you have a story idea we should look into? Tell us here.