Starting in March, the College Board will officially ban Internet-connected smart glasses — wearable mini computers built into eyeglasses that see, record and display information, and connect to Wi-Fi — during the SAT. For education leaders, the ban highlights how, even when the device in question is as small and inconspicuous as a student’s pair of glasses, a single new tool can change the risk landscape of assessments overnight.
According to Steve Addicott, chief operating officer at Caveon Test Security, a company that specializes in preventing cheating on exams, said the shift toward remote learning and the emergence of sophisticated AI tools have exposed the limitations of human-surveillance-heavy proctoring, especially when test content can spread online at a massive scale.
“I might say traditional proctoring hasn’t been all that effective, ever,” Addicott said. “It’s part of the mix. It’s important, we must have it, but we wouldn’t have our business if proctoring solved test fraud.”
Moreover, Addicott noted that most traditional proctoring models were not designed for the scale of modern digital education. Always-on monitoring or live proctoring — whether in person or via remote cameras — is expensive, difficult to staff consistently and prone to reviewer fatigue, he said, adding that asking teachers to act as “police officers” is often incongruent with their roles as educators.
“If you ask a teacher to go and walk around the room and look for people cheating, you’ve just changed the dynamic of that relationship,” he said.
As testing organizations respond to these pressures, Addicott said many, like the College Board, are reassessing whether increasing the number of proctors alone is enough to protect exam integrity. To him, doing that will require districts to analyze how exams are actually administered and taken.
“When a kid takes a test, the way they interact with the exam leaves a breadcrumb trail,” he said, adding that by applying statistical analysis to exam results, organizations can detect anomalies that human eyes might miss.
For example, Addicott noted it is “statistically improbable” for two students to have numerous identically incorrect answers on a 70-question test.
“It just doesn’t happen in normal test-taking,” he said. “We look at things like: Did the person finish a 90-minute exam in 12 minutes and get a 95 percent? That’s an indicator.”
LOCAL CHALLENGES AND ETHICAL TRAINING
In Utah, the focus remains on local control, where the responsibility for training proctors falls on local education agencies (LEAs). Scott Roskelley, test administration and data coordinator at the Utah State Board of Education, emphasized that traditional proctoring — where one or more professionals monitor a room of test-taking students with their own eyes — still works if educators are mindful of emerging risks.
“Good proctoring in Utah involves proctors moving around the classroom and paying attention,” Roskelley said. However, he acknowledged the technological advancement of students, mentioning that they have been known to smuggle in “burner phones” or use smart devices for language translation.
“We’ve had students that have been caught with two phones,” he said. “They turn one in, and they keep the other one in their pocket.”
Roskelley added that Utah requires anyone proctoring a test to undergo annual ethics training, which includes working through scenarios to determine what constitutes a breach versus a minor issue.
One of the most complex challenges in modern proctoring, according to Roskelley, is the balance between security and student needs. He pointed out that many students now use Internet-connected smart devices for legitimate medical reasons, such as monitoring blood sugar for diabetes.
“Those are smart devices that are attached to the Internet,” he said, noting that these situations are challenging because “it’s obviously something they need for their health and welfare, but it’s still an electronic device that can communicate.”
‘SECURE-BY-DESIGN’ TESTING
Addicott suggested that as technology evolves, test design must follow suit. He suggested the concept of randomly parallel testing — a method developed in the 1950s that is now feasible with modern technology — to prevent preknowledge and proxy test-taking. By using AI to create a larger pool of secure test formats, he said, organizations can move away from “fixed-form” tests that are easily pirated.
“If I give you a different test than I give the person sitting next to you, and those tests are equivalent in terms of difficulty, then the value of stealing that test goes way down,” Addicott explained.
In today’s digital era, both Addicott and Roskelley suggested that the prerogative of proctoring is shifting: It is no longer just about watching students, but about ensuring that every digital footprint confirms a score is a trustworthy measure of knowledge.
“The goal is to ensure that the test score is a valid reflection of what the student knows and can do,” Roskelley said.
