Image source, Getty Images
As large language models (LLMs) of artificial intelligence (AI) rapidly take over cognitive tasks, researchers warn of potential negative impacts as humans engage less in intellectual labor requiring thinking and effort.
Researcher Natalia Cosmina discovered that job applications she reviewed all appeared similar in style—sophisticated yet unrelated—especially the summaries, which did not align with the subject matter.
She later realized the applicants had used AI chatbots like ChatGPT, Google Gemini, and Claude to draft their documents.
Cosmina, who works at the Massachusetts Institute of Technology (MIT) in human-computer interaction, noted an increasing trend among students to forget course material over recent years.
She is concerned that growing dependence on LLMs may impair students’ cognitive abilities and intends to conduct further research on this subject.
Researchers worry that increasing trust in AI could erode language skills and mental capacities. Some studies suggest AI may contribute to a decline in human intellectual abilities.
The tools we use influence how we think. Since the internet arrived, we have increasingly sought answers through searching rather than deep thinking—a phenomenon called the ‘Google Effect.’ Some argue that the internet serves as an external memory system, easing cognitive load for the brain.
However, with increased reliance on LLMs and other AI, there is growing concern that intellectual effort may be reduced, weakening memory and problem-solving skills. AI can easily write poetry, give financial advice, and even act as a companion. Students, in particular, have begun offloading many intellectual tasks to AI.
Studies indicate these negative effects may hit children and adolescents hardest. Cosmina is eager to explore these issues more deeply.
Decline in Intellectual Effort
Cosmina and her MIT colleagues conducted an experiment in which 54 students were assigned to write a short essay. They were split into three groups. One group was allowed to use ChatGPT, another could use Google search but was prohibited from using AI-generated summaries, and the last group was instructed not to use any technology. Brainwave activity was monitored in all participants.
The essay topics were intentionally simple. Although the full study has not yet been published, Cosmina noted that the brains of those who relied on their own thinking were highly active. Students who only used search engines showed slightly more activity in visual regions, whereas those who used ChatGPT exhibited about 55% less brain activity.
“The brain wasn’t completely dormant, but the areas responsible for creativity and information processing were less active,” Cosmina explained.
ChatGPT use also negatively affected memory recall. Many students could not accurately cite facts present in their essays, and many denied ownership of the content generated by AI. Other studies have confirmed similar outcomes.
Image source, Getty Images
An international study confirmed that physicians who used AI tools to detect stomach cancer for three months found it challenging to work without those tools afterwards.
Cosmina warns that dependence on AI may reduce human creativity. Teachers reviewing ChatGPT-generated essays described them as “lacking personal talent,” unoriginal, and shallow. One teacher questioned whether these essays were collaboratively written by students sitting side-by-side.
While these studies demonstrate immediate effects of LLM use, the long-term consequences remain unclear.
Declining Cognitive Capacity
AI large language models can assist human thinking but should not be entirely relied upon, says computational neuroscientist and robotics expert Vivian Ming. However, she is concerned many people struggle to maintain this balance.
In her study, Ming asked University of California, Berkeley students to estimate real-world facts like oil prices. Many students were found to have generated responses directly using AI.
Measuring gamma wave brain activity, she observed lower dynamics among those students. Comparisons with other studies suggest decreasing gamma waves may correlate with future cognitive decline.
“This is very concerning,” Ming said. “Humans need to think deeply to remain capable. If we stop thinking, long-term cognitive capacities may deteriorate.”
While LLMs reduce intellectual effort, Ming emphasizes that mental exercise is critical to keep the brain healthy and to maintain balanced AI usage.
Less than 10% of the 54 students adopted alternate work methods—collecting and analyzing data solely from AI—which yielded better results and higher brain activity.
About two decades ago, Ming predicted that heavy dependence on Google Maps might increase dementia and memory decline rates over the next 20 to 30 years.
While precise data is lacking, studies show frequent GPS users tend to have poorer spatial memory. Another study linked this decline to increased Alzheimer’s disease risk.
Ming cautions that AI’s impact is not limited to creativity; it may also diminish intellectual capacity, raising dementia risks as brain activity diminishes.
As AI’s influence grows, addressing potential harms becomes crucial. Ming advocates for a “hybrid intelligence” model where humans and machines collaborate in rigorous intellectual effort.
Instead of relying wholly on AI, humans should think and learn independently before interacting with AI, a view Cosmina shares.
Ming recommends challenging one’s own thinking using what she terms the “Nemesis Prompt,” asking AI why one’s ideas may be flawed and how to improve them, enhancing thought quality and retention.
Another method is prioritizing “productive friction,” where AI is used to pose questions rather than provide answers.
Ultimately, shortcuts in intellectual activity should be avoided. Continuously challenging oneself is vital for long-term brain health, fostering creativity and cognitive strength.
For further materials on this topic, please visit our YouTube channel and social media platforms.
