Key Takeaways
- Researchers at the University of California, Irvine, studied how students worked before ChatGPT and their performance afterward.
- They found that AI lets students speed up homework in the moment but weakens long-term understanding.
- Still, students can use AI productively as a tutor while maintaining their critical thinking skills, according to a July study from Middlebury College.
Scientists are finding that passive reliance on chatbots can blunt critical thinking — unless you change how you use them.
A research team at the University of California, Irvine partnered with McGraw Hill to mine 10 years of usage data from Aleks, the company’s widely adopted adaptive math platform.
They pulled millions of problem‑solving sessions completed by students from fifth grade through college from 2015 to 2025, then treated the late‑2022 launch of ChatGPT as a natural breakpoint in the dataset.
The researchers studied how students worked in the pre‑ChatGPT era and how their behavior and performance changed after it launched.
To determine where AI was most likely to creep into students’ work, the team divided Aleks math items into two categories: word problems that could be easily copied into a chatbot and more visual, interactive graphing tasks that were difficult to paste into a chatbot.
The time students spent on word problems fell sharply after ChatGPT’s release even as their correctness rate went up. The authors said this pattern indicated that some students were using AI to do their work.
When students sat down for proctored exams without access to AI tools, they were 25% less likely to answer word problems correctly than comparable cohorts before ChatGPT’s arrival. Performance on graphing problems held steady.
Sina Rismanchian, the U.C. Irvine Ph.D. candidate who led the study, recently told The New York Times that AI allows students to offload the step-by-step thinking they need to solve math problems. AI enables them to speed up homework in the moment while weakening their understanding in the long-term.
Other studies clarify the relationship between AI use and critical thinking
In a separate multi‑institutional experiment cited by Rismanchian and his co‑authors, researchers looked at how adults tackled logic and reasoning problems with and without AI help.
As participants grew used to letting language models handle the heavy lifting and skipped the “productive struggle” of working through each step themselves, they became more likely to abandon challenging questions that were still within their ability to solve.
This mirrors broader findings that overreliance on AI makes people more prone to give up instead of persisting.
“Part of how AI is making us dumber” is “undercutting the development of learning how to think in younger people that now have an alternative,” Dr. Adam Green, a cognitive neuroscientist at Georgetown University, told the Times.
Green likens the brain’s impulse to avoid effort to the body’s drive to hang onto calories. “Whatever you could do that exerted the least calories but didn’t kill you was the best thing to do evolutionarily,” he told the outlet.
Still, students can use AI productively while maintaining their critical thinking skills. A July study from Middlebury College examined the habits of more than 200 undergraduate students. It found that students who used AI like a tutor had higher test scores and higher-quality essays than students who used it to draft their entire essays.
Key Takeaways
- Researchers at the University of California, Irvine, studied how students worked before ChatGPT and their performance afterward.
- They found that AI lets students speed up homework in the moment but weakens long-term understanding.
- Still, students can use AI productively as a tutor while maintaining their critical thinking skills, according to a July study from Middlebury College.
Scientists are finding that passive reliance on chatbots can blunt critical thinking — unless you change how you use them.
A research team at the University of California, Irvine partnered with McGraw Hill to mine 10 years of usage data from Aleks, the company’s widely adopted adaptive math platform.
They pulled millions of problem‑solving sessions completed by students from fifth grade through college from 2015 to 2025, then treated the late‑2022 launch of ChatGPT as a natural breakpoint in the dataset.