Proctoring systems developers are continually working on the improvement of AI algorithmic efficiency to monitor exams even better. For example, at Examus, we have increased the accuracy of face recognition technology to 99%. AI proctor performance, however, is not perfect yet. For instance, if there is a TV in the background with faces appearing on the screen, the cyberproctor may interpret this as the presence of other people in the room. Or a sudden lightning change, such as turning on and off a desk lamp, can be registered as a possible violation akin to replacing a student with another person.
Currently, we are working on bringing AI efficiency to a new level: the cyberproctor learns to recognize masked faces and correctly respond to lighting changes. We are also
developing an algorithm that is able to register violations by examining the students' long-term behavior from several events. For example, looking away every 3–5 seconds indicates cheating, and the cyberproctor will mark this as a violation. In a global sense, by developing proctoring systems, we are improving the artificial intelligence technologies impacting numberous spheres of our lives.