AI Detection Tools Not Reliable Enough to Prove Cheating, MSU Professor Says

Michigan State students fear false AI accusations, while a professor who doubts detection tools suggests oral exams and in-person tests as alternatives.

AI Tools
AI Detection Tools Not Reliable Enough to Prove Cheating, MSU Professor Says

A Michigan State University professor says oral exams and in-person writing could reduce the need for AI detection software. The idea comes as students report fears of being falsely accused of using artificial intelligence on their work.

Daniel Kramer teaches at the university’s James Madison College. He said he has used AI detection tools but does not trust them to show whether a student used AI.

“I don’t believe they’re reliable enough to say that you necessarily used AI, or you didn’t use AI,” Kramer told The State News. He now treats detection results as a reason to talk with a student, not as proof of misconduct.

Students Worry About False Flags

Economics freshman Poe Meh said social media posts about false accusations have made her nervous about turning in assignments. She said she has seen students accused of using AI when they had not. “So when you put in all of that hard work and then get accused of something like that, it doesn’t make you feel good,” Meh said.

International relations junior Matthew Horne also doubted the tools. “You’re relying on AI to tell you that I used AI when I actually didn’t use AI,” he said.

Horne said he uses AI to come up with ideas and to start assignments. He still believes students should own their work, saying thinking is “what your brain is supposed to be used for, not the computer.”

Kramer allows some AI use in his classes, including editing, spell-checking and polishing writing. He generally does not allow AI-written first or final drafts.

Oral Exams and Blue Books

Kramer said his college already focuses on critical thinking, writing and working with many sources. In-person blue book exams and heavy writing loads make it harder for students to rely fully on AI.

He suggested oral exams, or talks where students explain and defend their papers, as other options. “If they relied too heavily on AI, then quite likely they don’t have a grasp of that paper as they would if they had toiled and written the paper themselves,” he said.

Kramer said these methods take a lot of time for both professors and students. That is especially true in large classes.

Horne said such approaches could work in other programs but should fit each subject. He pointed to computer programming courses, which could use handwritten exams to test what students know.

Meh said handwritten work could ease some concerns but would not suit every course. She also noted that some professors use AI as part of their lessons.

She added that changing formats will not stop all cheating. “To be honest, even if we do it on paper, students who want to cheat will find a way,” Meh said.

Meh said the university should show more faith in students to reduce fear. “I think we just need to have more trust in our students,” she said.

Kramer said a survey he ran in one of his courses found that student views on AI differ across MSU colleges. He said professors must find “that fine line” between allowing some AI use and limiting students’ learning.

From our research desk
AI Jobs Automation Index
Which jobs are AI tools targeting most? We mapped 3,400+ AI tools to real job functions — with BLS employment & salary data.
Explore the index
maisiekooc

Written by

Maisie is a news writer at Agent Locker, covering the latest developments in artificial intelligence, emerging technology and the companies shaping the future.

Discover AI Agents