Dr. Jason Gibson noticed something strange when grading his students’ midterm exams. Across two summer history courses at Alcorn State University, Madagascar was an extremely popular reference. On its face, this wasn’t odd, but in this case, when the assignment asked to compare the Industrial Revolution with our digital age, phrases like “Madagascar floats sideways through the afternoon” made no sense.
The Madagascar reference wasn’t the result of some role the island nation played in the Industrial Revolution. Rather, it was the result of a trap that Gibson had laid while designing his exam. When he introduced the test, he hid the word “Madagascar” in white font so students wouldn’t see it—it was the same color as the background. Those who copied and pasted the exam language into an AI chatbot, however, prompted the artificial intelligence to come up with a response that referenced the hidden word.
So, how many of his students used an AI chatbot? “Thirty-two of my 35 students between two classes failed a portion of their midterm because they all used AI to generate their entire response,” he explained in a series of TikTok videos. “Apparently, they didn’t proofread it.”
The attitude toward Gibson’s gambit has been met with mixed reviews online. Many are praising it for its ingenuity, joking that the three students who did pass did actually use AI, but were savvy enough to proofread the exam before turning it in. Others point out that the scheme’s mileage will vary. “Just so you know, that won’t work much longer,” one commenter said. “The newer models don't have this issue and even tell the user when someone is trying to prompt inject them.”
After the exam, Gibson told the students why so many of them had failed and included proof of the white text via a screenshot. He invited anyone who had an issue with their grade to appeal.
AI and academic integrity are often at odds with one another. As students rely on things like ChatGPT and Claude to formulate their responses, critical analysis and research skills fall to the wayside. Educators are grappling with how to protect learning. Gibon’s invisible prompt shows just how prevalent AI usage is, even in settings where the point of being there is to enrich knowledge and learn life-long skills.
Thirty-two of 35 students in Dr. Jason Gibson’s summer history class failed their midterm exam. Why? Because they used AI to generate their entire response. Gibson explains how he caught them in a three-part TikTok series.
@iamjasongibson I can’t make this stuff up. #professor #highereducation #college #midterm ♬ Piano famous song Chopin Deep deep clear beauty – RYOpianoforte
@iamjasongibson #professor #highereducation #college #midterm #education ♬ original sound – Dr. Jason Gibson
@iamjasongibson Part 3 #professor #highereducation #college #midterm #education ♬ original sound – Dr. Jason Gibson
Sources: Professor's invisible prompt trap catches 32 students cheating on their midterm with AI; A Professor Created a Trap in His Midterm to Catch Students Using AI. Nearly All Made the Same Mistake
Related Articles:
Norway Announces Major Ban on Generative AI Use in Elementary School Classrooms
Harvard Grads Cheer for Ronny Chieng Who Says Their Generation’s Mission Is to “Destroy AI”
