Post Content The research revealed that even the shortest use of AI could have an alarmingly negative impact on a user’s ability to think. (Image: FreePik)
Artificial intelligence (AI) is making a lot of things easy. For instance, gone are the days when you would listen to a recording for hours to transcribe. And, if you are running out of ideas, an AI chatbot is at your disposal with a myriad of ideas, ready to brainstorm with you. Even though numerous research projects have shown the downsides of overreliance on AI tools, applications like ChatGPT, Claude and Gemini are witnessing rapid expansion of their user base. Amid the explosion of AI tools, new research warns about its impact on human minds.
Researchers from Carnegie Mellon University, Massachusetts Institute of Technology, the University of Oxford, and the University of California, Los Angeles, say that even 10 minutes of AI use may impact your ability to think. The study found that even the shortest use of AI could have an alarmingly negative impact on a user’s ability to think and solve problems.
“The takeaway is not that we should ban AI in education or workplaces,” Michiel Bakker, an assistant professor at MIT involved with the study, told Wired. According to Bakker, while AI can clearly help people to perform better in the moment, there is a need to be more careful about what kind of help AI provides and when.
Also Read | Anthropic turns to Elon Musk-owned SpaceX for AI compute as Claude demand surges
The team of researchers asked participants to solve various problems, including doing simple maths and reading comprehension, through an online portal that paid them for their work. The researchers conducted three tests, each involving hundreds of people. The team gave some of the participants AI assistants that could solve problems on their own. And, when these AI assistants were taken away suddenly, those relying on them were more likely to give up on a problem or make errors in their answers. The study found that excessive use of AI may boost productivity but at the cost of developing some foundational problem-solving skills.
The findings are from the paper titled ‘AI Assistance Reduces Persistence and Hurts Independent Performance’ published in arXiv. Apart from Bakker, the other authors are Grace Liu, Brian Christian, Tsvetomira Dumbalska, and Rachit Dubey. The key findings show how, after brief AI-assisted sessions (~10 minutes), participants were more likely to give up on problems and performed worse once the AI was removed compared to participants who never had AI assistance. The study shows this behaviour was seen among those participants who prompted AI to solve tasks for them directly. However, using AI for hints or clarifications did not lead to any significant impairment. Further, it shows that this is a general consequence of AI-assisted problem-solving, and it is not specific to any particular task.
Also Read | ‘AI brain fry’: Managing AI tools is mentally draining workers who want to quit, new study warns
According to the report in the Wired, Bakker was inspired to pursue how AI could be eroding people’s abilities following an essay on how AI would disempower humans over time. The essay argues that while AI risk scenarios usually portray sudden loss of human control, the incremental increase in AI capabilities without any coordinated power-seeking poses a substantial risk of eventual human disempowerment.
The essay and the studies also point to the need to focus on how AI can help people develop their mental abilities along with helping them with productivity. Bakker believes it is fundamentally a ‘cognitive question’ about persistence, learning, and how people react to challenges. “We wanted to take these broader concerns about long-term human-AI interaction and study them in a controlled experimental setting,” he told the publication.
Story continues below this ad
Also Read | In emergency room trials, Harvard study finds AI outperformed doctors with 67% accuracy
The MIT researcher feels that the outcome of the research was concerning. This is because a person’s willingness to continue with problem-solving is key to acquiring new skills; it also predicts their capacity to learn over time. This is in reference to the experiment where people quit problems the moment their AI assistants were taken away.
Bakker advocates for a different approach, one where AI sometimes focuses on a person’s learning over offering instant solutions for them. He claims that AI systems that give direct answers may have different long-term effects when compared with systems that ‘scaffold, coach, or challenge’ the user.
