A new study has sparked heated discussions: Does frequent use of AI writing really make the brain “lazy”? The experiment found that users who relied on AI had a significant decrease in neural activity and even had difficulty recalling what they had written. But the study also points out that this impact is not irreversible, and the key lies in how AI is used. Is it cognitive deterioration or efficiency gain? The answer may depend on whether you are still thinking actively.
Recently, a MIT research report on “the use of ChatGPT leads to cognitive debt accumulation” has caused quite a bit of controversy. Many media reported that the 216-page research report pointed out that the use of AI may reduce the brain’s intelligence. Sensational headlines such as “ChatGPT shrinks the brain by 47%” are emerging one after another, and many AI users feel anxious.
However, a closer look at the 206-page study reveals:The media’s interpretation has largely deviated from the core conclusions, and complex scientific research has been reduced to black-and-white assertions.In fact, MIT’s research does not support the simple assertion that “using AI makes people stupid”.
What this study wants to show is actually far deeper and farther than the media.
1. How did the media misread MIT’s conclusions?
Let’s briefly introduce MIT’s research “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task”. The study divided 54 subjects into three groups for writing tests: the pure human group relied on their own knowledge alone, the search engine group could use Google, and the AI-assisted group could use ChatGPT.
B-end product manager’s ability model and learning improvement
The first challenge faced by B-end product managers is how to correctly analyze and diagnose business problems. This is also the most difficult part, product design knowledge is basically not helpful for this part of the work, if you want to do a good job in business analysis and diagnosis, you must have a solid …
View details >
During the four-month trial cycle, college students in these three groups were asked to take 4 rounds of writing tests, which were equivalent to SAT-level essays, with a test time of 20 minutes and an interval of 1-2 weeks between each test.
MIT found through EEG monitoring that in the first three rounds of writing tests, the pure artificial group brain neural network was the most widely connected, and multiple brain regions such as memory extraction and logic integration were involved in collaborative work during writing. The search engine group has medium brain activity and relies on visual information management and screening capabilities, but still needs to integrate knowledge independently. The brain activity of the AI-assisted group was significantly reduced, the number of neural connections was reduced by 45%~55% compared to the pure artificial group, and 83.3% of AI users were unable to recall what they created a few minutes ago.
In the fourth round of cross-test, when the AI-assisted group wrote independently, although the brain activity rebounded, it was still much lower than that of the pure artificial group. When the pure artificial group used AI for the first time, brain activity increased instead of decreasing, and the quality of output was better than that of AI-dependent participants.The MIT experiment shows that the negative effects of AI are not irreversible, but long-term unconscious dependence on AI may create cognitive debt.Therefore, researchers encourage independent thinking before using AI.
However, some media outlets exaggerated the conclusions of the MIT study when disseminating them. The first is the misreading of the concept.Some reports of “brain atrophy by 47 percent,” but the MIT study measures changes in neural connectivity activity, not physical atrophy of brain structures. The study found that participants who relied on ChatGPT for writing had about 47% less neural coupling strength in their brains on tasks than the pure thinking group. This does not mean that the brain is “degenerated”, but indicates that certain cognitive areas of the brain are less active when performing specific tasks with the assistance of AI.
The second is the confusion of logical relationships.Cognitive debt is not an irreversible mental decline. The concept of “cognitive debt” proposed by the study refers to the fact that short-term reliance on AI may weaken long-term cognitive ability, similar to using it to abandon it. But this effect depends on how it is used. The study found that when the pure artificial group used AI for the first time, the brain activity did not decrease but increased, and the output quality was better than that of AI-dependent participants, while the AI-dependent experimental group had a significant decrease in performance once they were removed from the tool.Finally, the research conclusions are roughly simplified.The study does not negate the value of AI. The MIT paper clearly states that AI can be a cognitive enhancement tool, but only if users remain active thinkers. When people with strong cognitive abilities (high baseline cognition) used AI, neural connections were enhanced, and only those who relied on ChatGPT for a long time experienced temporary laziness in their brain thinking activities.
2. Controversy, does AI really corrode the brain?
After reading the report, we found that the MIT study not only did not directly conclude that “AI corrodes the brain”, but its research design was also not rigorous enough. We need to understand the conclusions it gives more objectively and prudently.1. The sample is representative of the limited.The number of participants was relatively small, initially 54, and only 18 completed the fourth phase. These people come from top academic institutions in the Boston area, belong to a typical WEIRD sample (Western, Educated, Industrialized, Rich and Democratic), and their cognitive habits, educational background, and technical literacy are above average. The conclusions drawn from this particular group cannot be directly generalized to a wider and more diverse range of ordinary users.
2. The experimental setting is different from the real environment.Completing an SAT-style philosophy essay in 20 minutes is a highly structured task. Simplifying the reality by equating the results produced by a single task with high pressure and time limits to a decline in overall cognitive ability. In real work, the process of people using AI is a non-linear, multi-step interaction with more time to reflect and adjust, and this difference may exaggerate the negative effects in the study.3. The measurement tool is not accurate enough.EEG (electroencephalography) technology has extremely high temporal resolution and can capture instantaneous changes in thinking, but has low spatial resolution and cannot detect deep structures such as the hippocampus, which are essential for the formation of long-term memory. In addition, EGG signals are also susceptible to external interference such as environmental electrical noise, which can affect the accuracy of the results. The researchers themselves acknowledged this in their paper and suggested that future studies should use techniques such as functional magnetic resonance imaging (fMRI) to obtain more comprehensive images of brain activity.
4. Correlation does not equal causation.In addition to being explained as laziness, the decrease in the number of neural connections may also be “cognitive offloading”. The decrease in brain activity may be due to the outsourcing of formatted, low-order cognitive tasks such as information retrieval to AI, freeing up valuable cognitive resources for higher-level strategic planning and critical research. Current experimental designs do not yet fully distinguish between these two distinct cognitive patterns.
In short, small-sample studies in a short period of time cannot draw one-size-fits-all conclusions, and to derive the conclusion that “AI causes cognitive decline” spread by the media, more rigorous and long-term longitudinal research is needed.
During the same period, some studies concluded that “AI tools help human brain thinking activity”, further supporting that “the use of AI does not lead to brain degeneration”.
In February 2025, a study from Microsoft Research and Carnegie Mellon University, “The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers”, A survey of 319 knowledge workers found that when users are confident in their expertise in the field, the use of AI stimulates more critical thinking.
A systematic review in the journal Computer and Education, “Higher-order thinking skills-oriented problem-based learning interventions in mathematics: A systematic literature review”, analyzed 69 relevant experimental studies between 2022 and 2024, It shows that the use of ChatGPT can improve students’ critical thinking, problem-solving, and creative thinking skills in higher-order thinking training. The study highlights that AI-assisted problem-based learning (PBL) can help students integrate concepts and reason logically more effectively.What these empirical studies want to convey is that we need to move beyond the simple binary of “AI is harmful or beneficial” and think about the more essential question: How can we use AI to make us better?
3. Embrace AI, not be domesticated by it
Historically, every technological revolution has caused panic about the degradation of human capabilities. The ancient Greek philosopher Socrates was worried that words would destroy human memory; The book “Entertainment to Death” warned that television leads to the decline of linear logical thinking; McLuhan once pointed out that computer systems may completely paralyze our central nervous system.
However, text, television, and computers do not make us stupid, but improve our skills. Writing does change the way of memory, but it does not eliminate the ability to remember, but creates a more complex knowledge system and continues to record valuable human civilization. Although television has changed the mode of information reception, it has cultivated new visual thinking ability and enriched our understanding of the world. Not only does computers not paralyze our nervous system, but they greatly expand the boundaries of human cognition, turning the world into a miniature global village.
The view that “AI will make people stupid” spread by the media often implies the misunderstanding of technological determinism and ignores the subjectivity of human beings in the use of technology.
It is undeniable that the brain is like a muscle, the more it is used, the stronger it becomes, and it deteriorates when it is not used. When writing by oneself, the brain has to go through the process of deep cognitive processing such as “conception, organization of language, expression, and modification”. When using AI, only the most shallow prompt input was done, and the logic and ins and outs behind the answers were not considered.
However, the impact of AI on cognition is not a one-way degradation or enhancement, but is highly dependent on usage and educational design.Just as there are people who use computers to learn courses around the world, and there are people who are addicted to playing games and suffer from mental illness, the key to the problem is how we use technology. If individuals rely on AI indiscriminately, passively, and comprehensively to complete all cognitive tasks, then in the long run, their ability to independently solve complex problems, think deeply, produce original results, and perform effective memory may be substantially reduced.
But the extent to which AI affects thinking varies from person to person and from person to person.Users who can actively use AI as an auxiliary tool rather than a complete replacement, maintain critical thinking during use, consciously engage in cognitive exercise and deep participation, may experience less negative impact, and may even be able to use AI to improve their cognitive efficiency and creativity.
If we can use AI brainstorming while adhering to the habit of critical thinking and deep memory writing, AI will not make us stupid, but will become a more intelligent search engine assistant.
So, returning to the MIT research report again, we will find that what it wants to express is:
Technology is neutral, and its impact depends on how it is applied.
However, to prevent addiction, the next time you’re ready to ask AI a question, ask yourself, “If AI didn’t exist, how would I solve this problem?” ”
Perhaps, the answer is in your own brain.
