‘It’s the illusion of learning’: How some educators say AI hurts students - The Detroit News

‘It’s the illusion of learning’: How some educators say AI hurts students

When ChatGPT burst onto the scene four years ago, some professors warned that students might use artificial intelligence to cheat.

When ChatGPT burst onto the scene four years ago, some professors warned that students might use artificial intelligence to cheat. Now many educators’ fears are far more profound: AI, it seems, is hurting their students’ ability to learn and think for themselves.

The technology is too new, and changing too fast, for definitive long-term data on its impact. But early evidence from some studies suggests that, when used without guardrails, large language models can damage students’ skill attainment. And some faculty members say they’re seeing troubling signs in the classroom: students who turn in polished essays but can’t answer questions about their ideas, students who fumble when trying to explain their code, students who have trouble with basic math skills.

Increasingly, educators grappling with students’ AI usage are asking fundamental questions about the nature of a college education, the skills of the future workforce and the collective knowledge of our society.

Eric Klopfer, a professor at MIT who co-chaired a committee studying AI in teaching, learning and research training, said in many cases students are turning in homework that is better. And their performance on exams is worse.

“It’s the illusion of learning,” he said. “It feels like you’re learning, it feels like you’re understanding things.” But when they need to apply that knowledge without the help of the LLM, students typically falter.

Fears surrounding artificial intelligence have been exploding in recent weeks, including warnings that the technology could pose an existential threat to humans. Multiple tech leaders have called for a pact to slow down the technology, while President Donald Trump has urged faster development.

It’s a watershed moment for higher education, MIT’s president told the campus community as the committee called for both immediate and long-term changes. There are early signals, the committee concluded, that overreliance on chatbots can have significant negative consequences - “diminishing critical thinking, weakening memory, eroding confidence, and undermining mastery.”

A December survey by the Rand think tank found growing numbers of students using AI for schoolwork, alongside concern about how that was affecting them. Two-thirds of students surveyed agreed that “the more students use AI for their schoolwork, the more it will harm their critical thinking skills.”

At Brown University, a committee studying generative AI found that most students have used LLMs in their studies - and they worry about the long-term impact on their thinking and learning.

Researchers are trying to catch up to understand behavior that has already become the default, said Luca Cian, a professor at the University of Virginia’s business school. “All students use AI,” Cian said.

Older studies have found that people don’t recall information as well when they have ready access to information through search engines or GPS maps, he said. LLMs, he said, are probably the most powerful cognitive off-loading mechanisms ever built.

One study this year, led by Grace Liu of Carnegie Mellon University, found that after just a few minutes of AI-assisted problem-solving, people who lost access to the AI performed worse on problems and gave up more quickly than those who had not used it.

Another study, published in 2025, measured brain activity while students wrote essays on their own, using a search engine or using an LLM to help. Researchers found the strongest brain connectivity in those doing the work on their own, and the weakest in those relying on AI - and the participants using an LLM struggled to quote from their essay after it was completed. Over four months, researchers reported, those using AI consistently underperformed on multiple levels.

Nataliya Kosmyna, a research scientist at MIT who was the lead author on that work, said that while longer-term studies are needed, it’s very concerning that changes are already apparent in such a short time period. Results are converging across different metrics, whether researchers are using brain scans, questionnaires or other means of assessing learning, she said. “There is a need and an urgency to do more of these studies.”

There’s a lot of debate about AI and security now, she said. “What about the brains of your children?”

In classrooms, the change is often palpable.

Shawna Dolansky, a professor in Carleton University’s College of the Humanities, was dismayed when her first-year students showed up with essay outlines that read like dissertation proposals. They were very polished, they often touched on topics that hadn’t been discussed in class, and some included quotations that weren’t in the books they had cited.

When she asked students about the ideas, they couldn’t answer. And these weren’t engineering majors trying to knock out a required English class; they were students who had signed up for a four-year humanities degree centered on Great Books.

“That’s a broken system,” Dolansky said.

At the University of California at Berkeley, Gireeja Ranade, a professor in the electrical engineering and computer sciences department, felt sad when she saw students failing the final exam in an optimization and linear algebra class; some of the problems were just like those they had been solving all semester for homework.

In previous years, LLMs weren’t advanced enough to solve the problems in the class. Now they can.

Some students, she said, were sincerely trying to just get a hint to help with their homework, but they ended up getting more assistance than they realized. “Sometimes they will confuse the idea of reading and understanding a proof with being able to do the proof themselves,” she said.

One of the key parts of being a student, she said, is being able to confront a problem and try things with the risk that they might be wrong. And when they had to do that on the exam, many couldn’t.

Dan Garcia, another professor in the department, also was surprised by the record high number of students who failed exams last spring - so high that the campus newspaper, the Daily Californian, wrote about it. In an introductory computer science class for students not majoring in the subject, he was shocked that more than half of the students apparently violated the school’s honor code and used AI for a take-home exam.

In an introductory class for students who were intending to major in computer science, he was surprised at how many students used AI for take-home projects, and then scored poorly on the in-person exam, apparently because they hadn’t learned the programming required.

“We’re seeing so many students who get a hundred percent by passing all the test cases, but when you ask them about their code, they have no idea, and they’re getting a zero or a one out of four” on an oral exam added to the class to combat overreliance on AI, he said. “They just don’t know their code because they didn’t write it.”

Garcia, an award-winning teacher, was also surprised by how few students were coming to class and to office hours last spring.

He has been hearing similar concerns from faculty globally. “I think this is a significant epidemic,” he said, “and I think educators like myself need to scramble.”

In many places, they are scrambling, with committees, university leaders and individual faculty members reevaluating how they teach, how they test and what is most important for students to learn.

Some professors are responding by eliminating or reducing technology, and adding oral exams or blue books. Some are flipping what they do in class and what they ask students to do outside of class. Some are designing AI tools that challenge their students to figure out answers or practice solving problems. And some are switching up syllabi.

“We as teachers need to think about how we structure our lessons for more in-class engagement, and think about motivation and what’s relevant and important and exciting to students,” said Erik Voss, an assistant professor at Columbia University’s Teachers College.

Dolansky upended what she had been doing for 30 years and stopped assigning take-home essays. It’s easy to see why a student would be tempted to not spend six hours on an assignment that their roommate finishes in seconds with AI, she said. She asks students to write in class, sometimes asking them to revise an earlier assignment.

And she made her classrooms tech-free, having students bring in physical copies of their books, take notes by hand in notebooks and talk through ideas in person. She assigns less reading but requires students to annotate their books in the margins, forcing them to engage with the text.

She said she thinks what she’s teaching students now is even more valuable in the wake of AI: “Here’s how humans think, and here’s how humans feel, and this is how humans express themselves.”