Mr. Williams’s case combines flawed technology with poor police work, illustrating how facial recognition can go awry.
The Shinola shoplifting occurred in October 2018. Katherine Johnston, an investigator at Mackinac Partners, a loss prevention firm, reviewed the store’s surveillance video and sent a copy to the Detroit police, according to their report.
Five months later, in March 2019, Jennifer Coulson, a digital image examiner for the Michigan State Police, uploaded a “probe image” — a still from the video, showing the man in the Cardinals cap — to the state’s facial recognition database. The system would have mapped the man’s face and searched for similar ones in a collection of 49 million photos.
The state’s technology is supplied for $5.5 million by a company called DataWorks Plus. Founded in South Carolina in 2000, the company first offered mug shot management software, said Todd Pastorini, a general manager. In 2005, the firm began to expand the product, adding face recognition tools developed by outside vendors.
When one of these subcontractors develops an algorithm for recognizing faces, DataWorks attempts to judge its effectiveness by running searches using low-quality images of individuals it knows are present in a system. “We’ve tested a lot of garbage out there,” Mr. Pastorini said. These checks, he added, are not “scientific” — DataWorks does not formally measure the systems’ accuracy or bias.
“We’ve become a pseudo-expert in the technology,” Mr. Pastorini said.
In Michigan, the DataWorks software used by the state police incorporates components developed by the Japanese tech giant NEC and by Rank One Computing, based in Colorado, according to Mr. Pastorini and a state police spokeswoman. In 2019, algorithms from both companies were included in a federal study of over 100 facial recognition systems that found they were biased, falsely identifying African-American and Asian faces 10 times to 100 times more than Caucasian faces.
Rank One’s chief executive, Brendan Klare, said the company had developed a new algorithm for NIST to review that “tightens the differences in accuracy between different demographic cohorts.”