"'Madagascar'? Caught red-handed"... U.S. university finds 32 of 35 students used an AI trap answer

[Financial News] A professor at a U.S. university has drawn attention after planting a hidden instruction in an exam paper that students could not see, in an effort to catch cheating with generative artificial intelligence (AI), only to confirm that most of them had simply used AI as-is.
According to Upworthy on the 25th (local time), Jason Gibson, a professor in the history department at Alcorn State University in Mississippi, recently posted a video on his TikTok account showing an experiment he conducted to check whether students used AI on a midterm exam, along with the results. The video has since drawn more than 1.6 million views.
Instead of using an AI detection program, Gibson chose to plant a trap directly in the exam questions.
He inserted a hidden instruction in an essay question on the Industrial Revolution using white text that matched the background color. It was normally invisible, but if students copied the question and pasted it into AI, the system would also recognize the hidden phrase.
The hidden sentence instructed students to include the word "Madagascar" somewhere in their answer, regardless of context.
The result was shocking.
Gibson said he found the word "Madagascar" in 32 of the 35 exam answers, appearing in contexts that had nothing to do with the Industrial Revolution. Some responses reportedly included bizarre phrases such as "Madagascar flowed sideways through the afternoon."
He added, "They failed the midterm question. The reason was that they not only used AI to write their answers, but also failed to review them."
However, he did not automatically treat all students as cheaters.
Through a class notice, Gibson shared screenshots showing where the hidden instruction had been placed and told any students who felt wrongly accused to raise an objection. Two students did so. One of them was able to have the penalty reversed by proving that white text had been invisible on his device because he was using dark mode.
He explained, "This experiment was not meant to punish students, but to show the reality of education in the AI era." He added, "AI has not been part of education for very long, so no one has all the answers. Teachers are simply trying different methods to protect academic integrity."
As the story spread, online users responded with comments such as, "All 35 students probably used AI, and the remaining three were just the ones who read their answers at least once," "It looks like in-person handwritten exams will make a comeback," and "Even if you use AI, shouldn't you at least review the answer?"
[email protected] Seo Yoon-kyung Reporter