Gabe Teninbaum, a professor at Suffolk University Law School, is calling on ExamSoft to fix a serious bug with its test-taking software: failure to recognize faces. It’s a problem that can delay test takers — or bar them from starting their exams altogether — and per reports, it disproportionately impacts people with dark skin tones.
ExamSoft’s software records students while they complete remote exams and monitors for signs of academic dishonesty. Teninbaum’s report addresses an ExamSoft feature called ExamID, which aims to verify that test-takers are who they say they are. The first time a student logs into their exam portal, they upload a photo of themselves (their “baseline image”); they’re then prompted to take another selfie before beginning future exams, which the software checks against their original photo.
Research has found that facial-recognition algorithms consistently make more errors in identifying Black faces than they do white ones. And while those studies didn’t focus on ExamSoft specifically, it doesn’t appear to be an exception. Back in September, multiple non-white exam-takers told the New York Times that the software couldn’t identify them due to “poor lighting” — a problem that Teninbaum, who has light skin, wasn’t able to replicate.