After failing two Uber facial recognition checks, a man was fired. Now, a UK labor union is suing the company for racial discrimination on his behalf. We’ve known for a while now that facial recognition AI have an inherent racial bias. Why is this still happening?
The UK’s Independent Worker’s Union has filed a lawsuit against Uber for its biased facial recognition program. After failing two ID checks, a driver was fired despite going through multiple lengths to address the false failures.
The driver, knowing that he was who he said he was, sent his facial ID only to be stonewalled and eventually fired for an error in the program. One would think that after two false negatives, Uber would manually verify the information of a driver’s ID. Facial recognition and AI can be finicky, and when it comes to someone’s job, flagrant dismissal of error would be discouraged. However, the driver was fired from the company without any human interaction whatsoever.
Credit: Youtube/TRT
The driver attempted to verify with an in-person representative of Uber who, while understanding the error, was “unable to do anything about it.” That’s comforting.
Face ID verification is a major part of Uber’s driver authentication security. The app mandates that drivers send multiple ID checks through the app. But when facial recognition is provably biased against people of color, the efficiency of AI systems can become a legal nightmare.
Facial recognition software has fundamental racial biases, the government has even straight-up admitted it. In a federal study by the National Institute of Standards and Technology (NIST), 200 facial recognition algorithms used by corporations were tested for accuracy. For mugshots, Asian, Black, and Native American people were misidentified sometimes 100 times more often than Caucasian people.
The study not only suggests a necessary rework of AI systems but that human bias can translate to “non-discriminant” computer programs.
Credit: Youtube/Hak5
Facial recognition is often presented as the next “big thing” in security technology for most corporations and governments. For example, the Dubai police force has invested in facial recognition-based robo-policing, in addition, some businesses are using facial recognition software to identify previous criminals/shoplifters. Seeing as many uses of facial recognition has to do with crime and policing in general, misidentification on the basis of race shouldn’t be swept under the rug.
The UK has struggled with the ethics of general facial recognition use in public spheres for a while now. The racial aspects of its errors don’t even scratch the surface of the controversy.
Faces aren’t like fingerprints. Have you ever ran into an old friend from high school, unable to put a name to a face? Or tried to tell the difference between the different Chrises of Hollywood? Even with the precision of AI, faces have less uniqueness than one would expect.
However, corporations likely enjoy the wall of impersonality that comes with the security system. It’s much easier to let an AI work its biased magic rather than a real representative.
As more and more corporations implement racially-biased facial recognition programs, we should expect to see more and more lawsuits pushing back.
