An international research team has documented a troubling cognitive cost to even brief reliance on artificial intelligence: people who use AI tools lose the ability to persist through difficult problems and perform worse when the assistance is removed, raising concerns about long-term erosion of human capability in an age of accelerating automation.
In controlled experiments on mathematical reasoning and reading comprehension, researchers from the University of Oxford, MIT, UCLA, and Carnegie Mellon found that participants who used AI for just 10 minutes performed worse and gave up more often when the tool was taken away compared with peers who received no help. The findings suggest that short-term productivity gains from AI may exact a hidden price: the gradual weakening of the cognitive stamina and skills that form the foundation for advanced learning.
The Cognitive Erosion Pattern
The research team described their findings as evidence of reduced persistence and impaired unassisted performance, warning that what appears to be a minor degradation in any single instance could compound into significant capability loss over time. The researchers employed the metaphor of a "boiling frog" to describe this process—a gradual, almost imperceptible erosion of human cognition as everyday AI use accelerates across workplaces, schools, and homes.
The concern extends beyond immediate task performance. The study's authors cautioned that skills such as fraction arithmetic and reading comprehension may appear easily delegable to tools, but conceptual mastery of them underpins higher-order capabilities including algebra and critical reasoning. This suggests that outsourcing foundational skills to AI could compromise the cognitive architecture required for advanced problem-solving and independent thought.
The Hidden Cost of Convenience
Grace Liu, a co-author from Carnegie Mellon University's Machine Learning Department, articulated the subtle nature of the problem. "The concern is about what cognitive scientists call 'desirable difficulties' - the productive struggle that builds skill over time. If AI routinely removes that struggle, people may get the right answer in the moment, but develop less robust independent capability," Liu said.
Liu emphasized that the issue is not crude—that AI makes people "dumber" in an obvious way. Rather, the mechanism is more insidious: by removing the difficulties that build durable competence, AI tools may quietly strip away the motivation and stamina required for long-term learning. Small degradations in persistence and problem-solving capacity, if they accumulate across years of AI use, could become difficult to reverse.
Research Gaps and Path Forward
Liu acknowledged that significant questions remain about the scale and scope of these effects. "How significant this effect is at scale, and across different contexts, needs more research," she said, noting that the current study provides evidence of the phenomenon but leaves open questions about its real-world magnitude in diverse settings.
Crucially, Liu framed the findings not as a reason to reject AI tools entirely, but as a call for intentional design and use practices. "It's not a reason to avoid AI, but it is a reason to design and use these tools carefully," she concluded, suggesting that the solution lies not in abandonment but in conscious choices about when and how to deploy AI assistance.
Why This Matters:
As artificial intelligence becomes embedded in educational systems, workplaces, and daily life, questions about its effect on human capability take on urgent social importance. The research suggests that without deliberate guardrails and thoughtful implementation, widespread AI adoption could create a two-tier cognitive landscape: those with access to tools but diminished independent problem-solving capacity, and those whose skills atrophy through non-use. This has implications for educational equity, workforce development, and the distribution of cognitive capability across society. If AI use erodes the foundational skills and persistence required for advanced learning, then access to AI—and decisions about when to use it—becomes a matter of public concern, not merely individual choice. The findings underscore the need for institutional frameworks, educational policy, and workplace practices that preserve human cognitive development even as these tools proliferate.