Lucy* (not her real name), a 29-year-old grocery delivery worker and mother of two in the US city of Houston, Texas, was under enormous financial strain when we interviewed her in May. She said her income had dropped steeply during the pandemic; she was on food stamps and barely making rent. In January, she applied for unemployment benefits, for which she seemed eligible after the US Congress extended them to gig workers affected by the pandemic.
But Lucy never received unemployment. On February 9, Texas denied her application: she had failed the unemployment agency’s automated identity check.
Lucy’s experience is not unique. According to a Motherboard investigation and local news outlets, workers in Texas and many other states are being denied benefits because of problems verifying their identity with the online identity platform, ID.me. To stem fraud, the platform uses facial verification – a type of facial recognition – to assess whether applicants’ selfies match their identity documents. Texas and at least twenty other states rely on ID.me to vet unemployment claims.
“[ID.me] had me take a picture and upload my driver’s license,” said Lucy, “But it said I wasn’t the person in the driver’s license.”
After the Motherboard report, ID.me CEO Blake Hall insisted their technology is “highly accurate (99.9%),” and errors are usually due to “image quality and lighting issues.” ID.me offers applicants a chance to fix errors via video chat, but many have reported long wait times. Hall promised to “keep video chat wait times down.”
As a last resort, applicants can appeal denials, but the process is lengthy and cumbersome. Lucy submitted appeals through the Texas unemployment agency’s website and called their helpline repeatedly, but to no avail: “I would be on hold for two hours before it cut me off and said they couldn’t take any more calls for the day.”
By March, she couldn’t afford rent. Since she was so busy looking for work, she also felt “horrible” she wasn’t spending enough time caring for her children, who were attending school remotely.
Addressing unemployment fraud is a legitimate goal. But deploying facial recognition should not come at the expense of the social protection needed to secure basic rights. At the very least, unemployment agencies should speed up appeals of identity verification errors and establish offline alternatives, such as in-person office visits.