This could leave us in a “boiling frog” scenario, in which the capabilities of our brains are progressively compromised as we lean on AI more heavily.
A team of researchers from the University of Oxford, MIT, UCLA and Carnegie Mellon said their research provides evidence for two consequences of using AI to help complete tasks – reduced persistence and impairment of unassisted performance.
The researchers asked people to carry out a variety of different tasks, including mathematical reasoning and reading comprehension.
They found that “after just 10 minutes of AI-assisted problem-solving, people who lost access to AI performed worse and gave up more frequently”.
Describing the gains seen from using AI as coming “at a heavy cognitive cost”, the team warned that their findings “raise urgent questions about the cumulative effects of daily AI use on human persistence and reasoning”.
They said: “These findings are particularly concerning because persistence is foundational to skill acquisition.
If AI routinely removes that struggle, people may develop less robust independent capability
“We caution that if such effects accumulate with sustained AI use, current AI systems – optimised only for short-term helpfulness – risk eroding the very human capabilities they are meant to support.”
While the effects may seem small in relation to society now, the team suggested a cumulative effect over years could lead to serious implications, undermining our ability to concentrate and learn.
If AI erodes the motivation and persistence required to drive long-term learning, the effects “will accumulate over years, and by the time they are visible they will be difficult to reverse”.
“This is analogous to the ‘boiling frog’ effect, where each incremental act feels costless, until the cumulative effect becomes overwhelming to address,” the team said.
Research co-author Grace Liu, from Carnegie Mellon University’s Machine Learning Department, said the study indicates that much more research about human interaction with AI is needed.
“The concern is about what cognitive scientists call ‘desirable difficulties’ – the productive struggle that builds skill over time,” she said.
“If AI routinely removes that struggle, people may get the right answer in the moment, but develop less robust independent capability.”