It’s been heralded as a revolutionary technology that will change everything about how we work and live our lives.
But now, top researchers in the US and UK are warning that artificial intelligence could be having an unintended consequence – it impairs the ability to think and problem solve.
After just 10 minutes of using an AI chatbot, researchers found that people were more likely to make mistakes and give up on tasks completely compared to those who never used the technology.
In a study, scientists from Carnegie Mellon, Oxford, MIT and UCLA, recruited 350 people and asked them to solve 15 fraction-based math problems.
Half were asked to solve the equations themselves, while the rest were given access to an AI assistant for the first 12 questions – which was then unexpectedly removed for the final three.
Participants with access to the AI fared better overall than those who did not have the tool for the first part of the experiment.
However, once the AI tool was removed, those same individuals performed much worse.
They had a 20 point lower average score for the final three questions and a twice as high skip rate compared to those who never used the technology.
Several large-scale studies have estimated that anywhere from seven to 15 percent of Americans use an AI chatbot at least once daily, equating to more than 30 million Americans.
Top researchers say they have found evidence that AI could undermine human cognition
Writing in the study, the researchers concluded: ‘We find that AI assistance improves immediate performance, but it comes at a heavy cognitive cost.
‘After just 10 minutes of AI-assisted problem-solving, people who lost access to the AI performed worse and gave up more frequently than those who never used it.’
They added: ‘These findings raise urgent questions about the cumulative effects of daily AI use on human persistence and reasoning. We caution that if such effects accumulate with sustained AI use, current AI systems… risk eroding the very human capabilities they are meant to support.’
Ever since Chat-GPT and other AI systems became popular in late 2022, tech entrepreneurs have been promising they will make the world better – but others caution they will also upend lives and replace millions of jobs.
Some have heralded the technology as revolutionary, saying it represents a similar disruption to the Industrial revolution, when, for the first time, societies found they had more people working in manufacturing than in farming.
Others, however, have sounded a more pessimistic note on AI, saying that it regularly makes mistakes, is a ‘useful idiot’ and is sycophantic to users.
Today, estimates suggest about 56 percent of US adults have used any AI tools, while 28 percent say they use them every week, and 13 percent use them daily.
In the study, which was published as a preprint – meaning it was not reviewed by other scientists – the researchers said people who used the AI likely found questions harder to answer because of cognitive offloading, when someone outsources the mental effort needed to complete a task.
In this, because someone has found it easier to complete a task using the technology, they may be more likely to skip that same task in the future if the technology isn’t available, rather than completing the task themselves.
The researchers added: ‘Human cognition has always been shape by external tools, from calculators to internet to GPS navigation.
‘Current AI systems, however, represent a new kind of cognitive scaffold: One that solves anything, rarely refuses to help, and delivers answers instantly.’
In the study, researchers also carried out a second experiment with another 600 individuals.
In this, all the individuals received three pretest problems to solve without the help of AI to boost their learning.
For the next questions, half answered them all independently, while the rest answered 12 using AI and then a further three after the AI was unexpectedly taken away.
Ever since Chat-GPT and other AI systems became popular in late 2022, tech entrepreneurs have been promising they will make the world better – but others caution they will also upend lives and replace millions of jobs
The test led to similar results to the previous study. But, when researchers analyzed how people used the technology, they found a startling difference.
The majority of AI users, 61 percent, used it just to get direct answers, and these individuals had the lowest scores and highest skip rates overall.
While 27 percent of participants sparred with the AI, using it to interrogate answers, and 12 percent said they refused to use it at all.
These individuals had higher scores than those who used AI directly. They also had higher scores than the group that was never offered the chance to use the AI.
The researchers concluded: ‘Just 10–15 minutes of AI interaction can result in significant impairments in independent performance and persistence – capacities that are foundational to life-long learning.
‘If brief exposure produces measurable erosion, the cumulative effects of daily AI use over months or years may be profound and difficult to reverse.’
