Remote proctoring vendors to senators: Trust us.
Asked by U.S. senators to address equity concerns being raised about their biometric products, executives at online proctoring firms say they have sufficient safeguards to protect students.
The firms were responding to a demand by senators late last year that the industry explain itself in the face of accusations that proctoring applications do not treat every biometrically monitored test taker equally.
The success rate of facial recognition systems when it comes to identifying people of color continues to be a sore point for vendors selling software to public and private organizations. Generally, the applications work best when dealing with middle-aged white males due to biased software training.
Biometric proctoring algorithms are created to go a step further, trying to spot people who are cheating on tests. Skeptics say all manner of innocent movements, including tics, can prompt erroneous alerts.
Those worried about privacy have also noted that students typically are taking tests at home during the pandemic, a setting in which people other than the student can appear in the background. Their biometric details — their faces — are recorded without their permission and held for a non-instant period.
An article in the trade publication Diverse: Issues in Higher Education, reported on the responses of leaders at ExamSoft, ProctorU and Proctorio, all of which amount to “we’ve got this covered.”
According to Diverse, organizations using ExamSoft’s Examplify test-session recording product can opt to employ a feature designed for face recognition and spotting the tells of cheats after a test.
Student identities can be confirmed manually later and, presumably, any suspicious behavior can be analyzed by a human as well.
Proctorio told the senators that it hired an independent AI and ethics consultancy last September to optimize its face detection. It reportedly is paying for independent research to see what if any bias hides in its software.
ProctorU claims to take a human-first approach to online proctoring. People watch students taking tests live along with an AI algorithm. The algorithm alerts the human monitor immediately if it see something dodgy happening.
Facial recognition works in reverse. Biometric AI does the identification, pulling in human eyes whenever it is unsure while performing authentication.