‘Cognitive Surrender’: Quick Solutions, Low Test Scores Show How AI Is Ruining Math Skills

Beginning in 2023, students began spending less time on word problems while continuing to spend about the same amount of time on graph problems. The gap widens every quarter. By the end of the academic year, in late 2025, the average time spent on word problems had dropped 31 percent among high school students and 27 percent among college students — from about four minutes per word problem to less than three. (Middle school students showed an average decline of 9 percent, and fifth graders showed none.)

The researchers believe that those averages are dragged down by some students who spend only seconds on word problems because they use AI to answer them.

A similar pattern emerged in college placement tests. When tests were taken unsupervised, students spent significantly less time on word problems after ChatGPT was released. During proctored tests, time spent on word problems has returned to historical norms.

But time is only part of the story. The most difficult finding is what happens in learning.

Many colleges allow incoming students to retake the placement tests after taking more math practice on the ALEKS, giving them a chance to qualify for higher-level courses. Before ChatGPT, that practice often paid off. After ChatGPT, students answered more word problems correctly during unsupervised practice sessions but performed worse on those types of problems when they later took the protocol placement test.

Historically, students have answered about 80 percent of these word problems correctly on supervised placement tests. After the introduction of ChatGPT, that dropped by nearly 60 percent — a 25 percent decrease in the likelihood of answering the word problem correctly.

Performance on the graph problems, in contrast, did not decrease.

After the ChatGPT was released, students performed worse on word problems (AI-susceptible) during proctored tests, but answered more word problems correctly in nonproctored settings.

If students’ math skills had been deteriorating due to epidemic learning loss, poor high school preparation or digital disruption, the graph’s performance should have been worse. It didn’t happen.

The study cannot definitively prove that students were using AI. The researchers couldn’t see what else was happening on the students’ screens other than ALEKS. But it is difficult to think of another explanation. Changes appeared only in problems that were easy to give to the AI, disappeared under supervision and grew slowly in about three years.

“What scares me is that it’s not just about the word problems,” Rismanchian said. “This dedication of the mind can continue through writing, science, everything.”

The paper, “Faster Completion, Less Learning,” was released in June 2026 as a working paper and has not yet been peer-reviewed. As with any single study, it does not address the questions of how many students are using AI in their schoolwork, or whether it is harming learning and to what extent. But it also includes a growing body of evidence that generative AI is causing students to skip the brain activity that leads to learning, and that this “brain surrender” is becoming commonplace.

A randomized experiment in Turkey found that high school students who used AI to help them learn math ended up reading less than students who practiced without it. Anthropic, the maker of Claude, reported separately that more college students seem to be using AI to get answers and outsource cognitive work. Rismanchian’s previous study, released in March 2026, documented troubling patterns of AI use in short essay writing among undergraduates at a major research university in California.

That doesn’t mean AI always underestimates learning. Carefully designed AI tutors improve students’ success on controlled tests by asking questions, personalized instruction and holding answers until students think their way through a problem. But using AI in this way should increase the time students spend on the problem, Rismanchian said. ALEKS data shows the opposite.

Rismanchian doesn’t believe the answer is simply preventing AI. Instead, he says, students need to value learning enough to resist the temptation to find it out.

A recent RAND survey suggests that many already recognize the danger in their brains. Students report that they are worried that AI is weakening their critical thinking skills while many of them admit that they use it for school work.

The students are completely innocent. As many professors have warned students not to use AI to complete class work, universities themselves have embraced the technology, often giving students free access to premium chatbots.

“I think we need to communicate to students that you should value your learning,” Rismanchian said. “If ChatGPT does it for you, you haven’t read it yet.”

Rismanchian understands the temptation.

An international student, Rismanchian started using ChatGPT to help polish the English in his papers. The ideas were still his. But after a few months, he said, he noticed something that didn’t make him happy.

“I realized that I couldn’t write anymore,” he said. “I was losing my ability to write.”

So he stopped using AI to write.

You still use it to write code.