Mark Zuckerberg famously tapes over the lens of his webcam when he’s not Zooming. Even if they don’t block the lens, many people feel queasy about webcam video’s obtrusiveness. Students taking online proctored exams with webcams are no different. In fact, they may be more concerned about appearing on a webcam because many students have already given up privacy and control in living with roommates, and proctored exams via webcam are another intrusion on their limited personal space. Even when they accept the necessity of proctoring tests to ensure that they are fair, students question what the remote monitor can see, what happens to the data, and how it is used.
Part of the discomfort comes from the novelty of online proctored exams. For students who’ve been working the whole semester remotely, accustomed to only open-book quizzes, taking an online exam with webcam is new and scary. Familiarizing students with the technology and procedures of proctoring helps to lower their anxiety around them.
Dylan Singh, a member of Rosalyn’s student advisory board and a junior at the University of Southern California, says:
[The sudden use of online proctoring is] the biggest anxiety factor because, when our teacher says, all right, this final is going to be using proctoring technology, we think to ourselves: now the emphasis isn’t on us showing our learning, it’s on not getting caught.
One strategy schools use to lessen online proctoring’s scariness is familiarizing students with it over the semester with lower-stakes exams. Robust proctoring systems for online exams allow you to adjust the proctoring level to the stakes of the test. For example, professors can dial down the proctoring on quizzes by omitting the remote webcam monitoring. Professors can still impose time limits and lock down the browser to outside websites or files during the exam. When the stakes are higher during a final or midterm, they can dial the proctoring up to include webcam, keystroke, and microphone monitoring.
For administrators implementing a proctoring solution, being transparent with instructors and students about how the proctoring works goes a long way toward making them more comfortable with the technology.
Proctoring should be about deterrence. Schools don’t want to catch cheaters so much as they want students to play fair in the first place. If students believe that it is difficult to cheat on exams and that very few people attempt to cheat, they are far less likely to do so themselves. This idea was summed up by a member of Rosalyn's student advisory board, Jessica Ramses who is a junior at the University of Pennsylvania:
If the system doesn't proctor well enough, people in the class are going to cheat. They will be getting a better grade than you, despite their working less and getting additional help.
Conversely, students are afraid that they’ll be flagged by proctoring software for something completely innocuous, such as their roommate popping into their room or having to use the restroom. Many students echo the same questions: Does it catch cheaters? Does it falsely flag innocent things? These are two sides of the same coin. The answer to both is primarily determined by who you choose for an online exam proctoring vendor.
Remote invigilation systems that flag every behavior, innocent or not, erode confidence in the technology. Most proctoring systems use off-the-shelf AI models in proctoring online exams. These general-purpose AI models for computer vision are robust in that they can identify a wide range of objects in the webcam datastreams, but they are not precise at flagging the kinds of behaviors that would be violations in a test session.
Advanced systems have purpose-built AIs specifically trained on test session data. These systems constantly improve as the developers add session data from an increasingly diverse population. Growing the dataset with sessions from people with a range of skin tones, cultural backgrounds, and neurotypes expands the set of events that the system considers normal and ordinary and improves its ability to find behaviors that indicate a test violation.
Some online proctoring systems combine the efficiency and sensitivity of automated AI proctoring with the judgment and discretion of a human to review and verify the machine’s flags. Human-in-the-loop AI proctoring is the right way to do it. No one wants a robot to decide whether a student passes or fails a test. When a proctoring report for a test session consists of incidents that a human confirmed, it goes a long way towards showing that online proctoring systems work as advertised.
Online proctored exams with a webcam can be more comfortable for students by making a positive student experience a priority. Rosalyn’s focus on the student experience guides our technology development and supports educational institutions in our deployments. To sum up:
Finding an online exam proctoring vendor using webcams that espouses these principles is half the battle in making students comfortable with it. The other half is diligently implementing the solution in consultation with the student and faculty users. Together, these efforts will help make your online assessment initiative successful.
Welcome to the future of cheating, where AI isn't just an ally but an accomplice. In CheatCode 2.0, we're delving into the unexpected frontier of academic dishonesty—where the machines that are programmed to help us learn can also be hijacked to game the system. Get ready for a journey through the intricate maze of ethical dilemmas and technological advancements, as we unravel why AI might be the newest threat to academic integrity and what Rosalyn.ai is doing to level the playing field