Professors Reinvent Exams to Turn AI Into Learning Tool

When Nikita Bezrukov started seeing essays where half the cited research didn't exist, he knew something had to change. The MIT linguistics professor was witnessing what educators globally now face: students using AI tools like ChatGPT on coursework, sometimes breaking the rules, often leaving telltale signs like hallucinated references.

The numbers tell the story. A 2024 survey of over 1,000 UK undergraduates found 94% use AI tools for assessed work, while a US study covering 95,000 students at 20 universities estimated 9% knowingly broke rules by using AI on assignments.

But instead of fighting an unwinnable battle, professors are getting creative. They're redesigning tests to make AI part of the learning experience while exposing students who misuse it.

Nicholas Mattei at Tulane University now assigns students to compare how three different AI models summarize sources, then critique the results. Students earn extra points for catching when the AI invents fake references or creates misleading infographics.

Daniel Silver at the University of Toronto ditched traditional essays about economic theory. Now his students build virtual marketplaces using AI agents, then analyze the transcripts those agents produce to write their papers—a task that instantly reveals who asked AI to do everything in one shot.

Computer scientist Ruben Verborgh takes a different approach: he asks open-ended questions that AI can't answer well. "Why is this website slow?" has multiple valid answers requiring sound methodology, something chatbots consistently fail to provide.

Some professors are embracing transparency. Bezrukov now requires students to submit work through Google Docs so he can review revision histories and see exactly how AI was used in the process. He doesn't penalize students using AI to polish grammar, only those generating entire texts.

Others moved assessments back into classrooms. Kristina Ruiz-Mesa at California State University replaced take-home infographic assignments with in-class pitches and presentations after seeing too much AI-generated material.

Why This Inspires

These educators aren't just solving a cheating problem. They're preparing students for a future where AI is everywhere, teaching them to use it wisely rather than pretending it doesn't exist.

Yanjun Shen at Chang'an University in China asks PhD students to share their AI interaction logs, not to punish them, but to ensure the tools supported their thinking rather than replaced it. The goal is helping students expand their knowledge through AI, not surrender their intellectual work to it.

As Mattei admits, nobody has the perfect answer yet, and things are changing rapidly. But that's exactly why this matters—educators are adapting in real time, turning a challenge into an opportunity to teach critical thinking for the AI age.