Last spring, Shiri Melumad, a professor at the Wharton School of the University of Pennsylvania, gave a group of 250 people a simple writing assignment: Share advice with a friend on how to lead a healthier lifestyle. To come up with tips, some were allowed to use a traditional Google search, while others could rely only on summaries of information generated automatically with Google’s artificial intelligence.
The people using A.I.-generated summaries wrote advice that was generic, obvious and largely unhelpful — eat healthy foods, stay hydrated and get lots of sleep! The people who found information with a traditional Google web search shared more nuanced advice about focusing on the various pillars of wellness, including physical, mental and emotional health.
The tech industry tells us that chatbots and new A.I. search tools will supercharge the way we learn and thrive, and that anyone who ignores the technology risks being left behind. But Dr. Melumad’s experiment, like other academic studies published so far on A.I.’s effects on the brain, found that people who rely heavily on chatbots and A.I. search tools for tasks like writing essays and research are generally performing worse than people who don’t use them.
“I’m pretty frightened, to be frank,” Dr. Melumad said. “I’m worried about younger folks not knowing how to conduct a traditional Google search.”
Welcome to the era of “brain rot,” the slang term to describe a deteriorated mental state from engaging with low-quality internet content. When Oxford University Press, the publisher of the Oxford English Dictionary, named brain rot the word of the year in 2024, the definition referred to how social media apps like TikTok and Instagram had people hooked on short videos, turning their brains into mush.
Whether technology makes people dumber is a question as old as technology itself. Socrates faulted the invention of writing for weakening human memory. As recently as 2008, many years before the arrival of A.I.-generated web summaries, The Atlantic published an essay titled, “Is Google Making Us Stupid?” Those concerns turned out to be overblown.
But the growing wariness in academia of the impact of A.I. on learning (on top of older concerns about the distracting nature of social media apps) is troubling news for a country whose performance in reading comprehension is already in steep decline.
This year, reading scores among children, including eighth graders and high school seniors, hit new lows. The results, gathered from the National Assessment of Educational Progress, long regarded as the nation’s most reliable, gold-standard exam, were the first of their kind to be published since the Covid-19 pandemic disrupted education and drove up screen time among youths.
Researchers worry that evidence is mounting of a potent link between lower cognitive performance and A.I and social media. In addition to recent studies that found a correlation between the use of A.I. tools and cognitive decline, a new study led by pediatricians found that social media use was associated with poorer performance among children taking reading, memory and language tests.
Here’s a summary on the research so far, and how to use A.I. in a way that boosts — rather than rots — the brain.
When we write with ChatGPT, are we even writing?
The most high-profile study this year about A.I.’s effects on the brain came out of the Massachusetts Institute of Technology, where researchers sought to understand how tools like OpenAI’s ChatGPT could affect how people write. The study, which involved 54 college students, had a small sample size, but the results raised important questions about whether A.I. could stifle people’s abilities to learn.
(The New York Times has sued OpenAI and its partner, Microsoft, claiming copyright infringement of news content related to A.I. systems. The two companies have denied those claims.)
For part of the study, students were asked to write an essay ranging from 500 to 1,000 words, and they were divided into different groups: One group could write with the help of ChatGPT, a second group could look up information only with a traditional Google search, and a third group could rely only on their brains to compose their assignment.
The students wore sensors that measured electrical activity in their brains. The ChatGPT users showed the lowest brain activity, which was unsurprising since they were letting the A.I. chatbot do the work.
But the most striking revelation arose after the students finished the writing exercise. One minute after completing their essays, the students were asked to quote any part of their essay. The vast majority of ChatGPT users (83 percent) could not recall a single sentence.
In contrast, the students using Google’s search engine could quote some parts, and the students who relied on no tech could recite lots of lines, with some even quoting almost the entirety of their essays verbatim.
“It has been one minute, and you really cannot say anything?” said Nataliya Kosmyna, the research scientist at M.I.T. Media Lab who led the study, about the ChatGPT users. “If you don’t remember what you wrote, you don’t feel ownership. Do you even care?”
Though the study focused on essay writing, Dr. Kosmyna said she worried about the implications for people using A.I. chatbots in fields where retention is essential, like a pilot studying to get a license. More research urgently needs to be done, she said, on how A.I. affects people’s ability to hold on to information.
Social media may be linked to lower reading scores.
Over the last two years, schools in states like New York, Indiana, Louisiana and Florida have raced to ban cellphones from classrooms, citing concerns that students were distracting themselves with social media apps like TikTok and Instagram. Lending credence to the bans, a study published last month found a potent link between social media use and poorer cognitive performance.
Last month, the medical journal JAMA published a study conducted by the University of California, San Francisco. Dr. Jason Nagata, a pediatrician who led the study, and his colleagues looked at data from ABCD, for Adolescent Brain Cognitive Development, a research project that followed more than 6,500 youths from age 9 to 13 from 2016 to 2018.
All the children were surveyed once a year on how much time they used social media. Every other year, they took several tests. For example, a visual vocabulary test involved correctly matching pictures to words they heard.
The data showed that children who reported using a low amount of social media (one hour a day) to a high amount (at least three hours a day) scored significantly lower on reading, memory and vocabulary tests than children who reported using no social media.
As for why social media apps like TikTok and Instagram would harm test scores, the only safe conclusion is that every hour a child spends scrolling through the apps takes time away from more enriching activities like reading and sleeping, Dr. Nagata said.
What are some healthier ways to use social media and A.I.?
Despite findings of a correlation between social media use and cognitive decline, it would be difficult to recommend an ideal amount of screen time for youths, because lots of children spend time in front of screens doing things unrelated to social media, like watching TV shows, Dr. Nagata said.
Instead, he suggested that parents enforce screen-free zones, prohibiting phone use in areas like the bedroom and dinner table so that children can stay focused on their studies, sleep and mealtimes.
Meta did not respond to a request for comment. A TikTok spokeswoman referred to a webpage with instructions to set up Time Away, a tool for parents to create schedules for when their teenagers are allowed to use TikTok.
As for A.I. chatbots, there was an interesting wrinkle in the M.I.T. study that presented a possible solution on how people could best use chatbots to learn and write.
Eventually, the groups in that study swapped roles: The people who relied only on their brains to write got to use ChatGPT, and the people who had relied on ChatGPT could use only their brains. All the students wrote essays on the same topics they had chosen before.
The students who had originally relied only on their brains recorded the highest brain activity once they were allowed to use ChatGPT. The students who had initially used ChatGPT, on the other hand, were never on a par with the former group when they were restricted to using their brains, Dr. Kosmyna said.
That suggests that people who are eager to use chatbots for writing and learning should consider starting the process on their own before turning to the A.I. tools later in the process for revisions, similar to math students using calculators to solve problems only after they have used pencil and paper to learn the formulas and equations. Both Google and OpenAI declined to comment.
Dr. Melumad, the Wharton professor who led the earlier study involving A.I. search tools, said the problem with those tools was that they transformed what was once an active process in your brain — perusing through links and clicking on a credible source to read — into a passive one by automating all of that.
So perhaps the key to using A.I. in a healthier way, she said, is to try to be more mindful in how we use them. Rather than ask a chatbot to do all the research on a broad topic, Dr. Melumad said, use it as a part of your research process to answer small questions, such as looking up historical dates. But for deeper learning of a subject, consider reading a book.
Brian X. Chen is the lead consumer technology writer for The Times. He reviews products and writes Tech Fix, a column about the social implications of the tech we use.
The post How A.I. and Social Media Contribute to ‘Brain Rot’ appeared first on New York Times.
