This article, “The Best Algorithms Struggle to Recognize Black Faces Equally” by Tom Simonite is about facial recognition software. Facial recognition software is most often used as security measures in order to identify criminals. This kind of technology is useful because it helps identify people who may want to harm us, such as terrorists or other kinds of criminals. The article uses the example that the Idemia’s algorithm scans faces by the million and is used to check the faces of cruise ship passengers landing in the U.S. against Customs and Border Protection records. It is also used by the FBI to check mugshots. In addition, the police use it to find suspects more quickly.

The problem is, that the National Institute of Standards and Technology indicated that Idemia’s latest algorithms were more likely to mix up black women’s faces than those of white women or white or black men. Idemia’s algorithms falsely matched black women’s faces 10 times more frequently than white woman’s faces. In a report that tested the code of more than 50 companies, many of the top performers show similar performance gaps to Idemia’s error rate for black and white women and these facial recognition algorithms consistently perform less well for women than men which are attributed partly because of the use of makeup. These errors could lead to racial disparities in policing.

One way to fix the problem could be to restrict or ban the use of the technology until its performance issues are resolved or greatly improved. Officers could also check suggested matches carefully and seek corroborating evidence before taking action against an individual which some departments have been doing and have no issues with lawsuits, cases, or complaints. Another possible solution could be to have datasets that include people with darker skin since most datasets have 81% or more people with lighter skin. In my opinion, using this technology in police departments is fine as long as officers aren’t only relying on the technology and do what the officers in a previous example did. In other places though, I would be more hesitant in using them and possibly restrict their use until the technology improves.