By August 2021, eight months had passed since 62-year-old Michael Feldman was assaulted at a train station in Pagedale, Missouri. So when cops turned up with a lineup of men, Feldman felt “cornered.” Not only had his memory faded, but he’d sustained a brain injury.Police goaded Feldman into selecting 29-year-old Chris Gatlin – the only “evidence” to have Gatlin arrested.
Between police pressure, trauma, and the fallibility of human memory, witness identification is already notoriously unreliable. So was the software that selected Gatlin in the first place.
An AI facial recognition program matched his mugshot from a prior traffic violation to a poor-quality image of Feldman’s attacker. As a result, Gatlin spent 16 months in jail. This could happen to anyone, no matter what we do or how “innocent” we think we are.
After 16 months in jail and cleared charges last March, Gatlin moved to sue in January. But this story doesn’t end with him.
What will the future of anti-Black policing be with AI tools in its arsenal? Before AI, other scientific tools have proved to be rife with errors, from hair analysis and arson investigations to fingerprinting and drug tests. Gatlin’s story reinforces a need to think critically about them all. The criminal legal system may pick up shiny new toys, but it’s always been the same old dog playing the same anti-Black tricks.