DMV's AI System Says Woman Doesn't Have a Human Face

"When it comes to inclusion, artificial intelligence doesn't have the best track record. As vast algorithms trained on troves of data scraped from the internet, AI models are inherently predisposed to reproducing human social biases. It's no surprise, then, that AI has a real knack for discrimination, often exacerbating prejudice on the basis of race, gender, and sexuality."

"Gardiner, who lives with Freeman-Sheldon syndrome - a rare genetic disorder affecting muscles around the face, particularly the mouth - says that one by one, her photos were all rejected by the DMV's ID software. It became a spectacle, she told Wired. "Everyone's watching," she said. " They're taking more photos." "It was humiliating and weird," she continued. "Here's this machine telling me that I don't have a human face.""

Artificial intelligence models trained on internet data routinely reproduce human social biases and can exacerbate prejudice across race, gender, sexuality, and disability. Automated systems used for identification and content moderation often fail to recognize people with visible differences, producing discriminatory outcomes. A Connecticut woman living with Freeman–Sheldon syndrome experienced repeated photo rejections from a DMV ID verification program, causing public humiliation and denial of service. Visible differences include scars, burns, cancer-related changes, craniofacial conditions, hair loss, vitiligo, and inherited disorders. Many people with visible differences report that biometric and social media AI tools complicate daily life and access to services.

Read at Futurism