Is your brain really being "corroded" by AI?

Recently, a research report from MIT titled "Using ChatGPT Leads to the Accumulation of Cognitive Debt" has sparked quite a bit of controversy.

Many media reports claim that the 216 - page research report points out that using AI may cause a decline in brain intelligence. Sensational headlines like "ChatGPT Shrinks the Brain by 47%" are emerging one after another, leaving many AI users feeling anxious.

However, a careful reading of this 206 - page research report reveals that the media's interpretation largely deviates from the core conclusion, and the complex scientific research has been simplified into a black - and - white assertion.

In fact, the MIT study does not support the simple claim that "using AI makes people stupid."

What this study intends to illustrate is actually much deeper and more far - reaching than what the media has spread.

01 How Did the Media Misinterpret MIT's Conclusion?

Let's first briefly introduce MIT's study, "Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task."

This study divided 54 subjects into three groups for a writing test: the manual group relied solely on their own knowledge, the search - engine group could use Google, and the AI - assisted group could use ChatGPT.

During the four - month test period, the college students in these three groups were required to take four rounds of writing tests. The difficulty level was equivalent to that of an SAT (American college entrance exam) essay, with each test lasting 20 minutes and an interval of 1 - 2 weeks between each test.

Through electroencephalogram (EEG) monitoring, MIT found that in the first three rounds of writing tests, the manual group had the most extensive neural network connections in the brain. When writing, multiple brain regions such as those involved in memory retrieval and logical integration worked together. The search - engine group had a moderate level of brain activity, relying on visual information management and screening abilities but still needing to integrate knowledge independently. The brain activity of the AI - assisted group decreased significantly, with the number of neural connections 45% - 55% less than that of the manual group. Moreover, 83.3% of AI users were unable to recall the content they had created just a few minutes ago.

In the fourth round of cross - testing, when the AI - assisted group wrote independently, their brain activity increased but was still far lower than that of the manual group. When the manual group used AI for the first time, their brain activity increased instead of decreasing, and the quality of their output was also better than that of participants who relied on AI.

The MIT experiment shows that the negative impact of AI is not irreversible, but long - term unconscious reliance on AI may lead to cognitive debt. Therefore, the researchers encourage a thinking mode of independent thinking first and then using AI.

However, some media exaggerated the conclusions of the MIT study when spreading the news.

Firstly, there was a misinterpretation of the concept. Some reports claimed that "the brain shrank by 47%," but the MIT study measured the change in the activity of neural connections, not the physical shrinkage of the brain structure. The study found that the neural coupling strength of participants who relied on ChatGPT for writing decreased by about 47% compared to the pure - thinking group during the task. This does not mean that the brain is "degenerating," but rather indicates that under AI assistance, the activity of certain cognitive regions in the brain when performing specific tasks decreases.

Secondly, there was a confusion of logical relationships. Cognitive debt is not an irreversible decline in intelligence. The concept of "cognitive debt" proposed in the study refers to the fact that short - term reliance on AI may weaken long - term cognitive abilities, similar to the principle of "use it or lose it." However, this impact depends on the way of use. The study found that when the manual group used AI for the first time, their brain activity increased instead of decreasing, and the quality of their output was also better than that of participants who relied on AI. Once the experimental group that relied on AI was separated from the tool, their performance declined significantly.

Finally, the research conclusions were roughly simplified. The study did not deny the value of AI. The MIT paper clearly states that AI can be a cognitive enhancement tool, but only if users maintain active thinking. When people with relatively strong cognitive abilities (high - baseline cognizers) use AI, their neural connections are actually enhanced. Only those who rely on ChatGPT for a long time will experience a brief slack in brain thinking activities.

02 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 also had some flaws in its research design. We need to understand its conclusions more objectively and prudently.

1. Limited sample representativeness. The number of participants was relatively small. Initially, there were 54 people, and only 18 completed the fourth stage. All of them were from top academic institutions in the Boston area, belonging to the typical WEIRD sample (Western, Educated, Industrialized, Rich and Democratic). Their cognitive habits, educational backgrounds, and technological literacy were all higher than the average level. The conclusions drawn from this special group cannot be directly extended to a wider and more diverse group of ordinary users.

2. The experimental setup was different from the real environment. Completing a SAT - style philosophical essay within 20 minutes is a highly structured task. Equating the results of a single high - pressure and time - limited task with a decline in overall cognitive ability simplifies the real - world situation. In real work, the process of people using AI is a non - linear, multi - step interaction, with more time for reflection and adjustment. This difference may exaggerate the negative effects in the study.

3. The measurement tool was not precise enough. EEG (electroencephalogram) technology has extremely high temporal resolution and can capture instantaneous changes in thinking, but its spatial resolution is low, and it cannot detect deep - seated structures such as the hippocampus, which is crucial for forming long - term memories. Moreover, the EGG signal is also easily affected by external interference such as environmental electrical noise, which may affect the accuracy of the results. The researcher himself also admitted this in the paper and suggested that future research should use techniques such as functional magnetic resonance imaging (fMRI) to obtain a more comprehensive image of brain activity.

4. Correlation does not equal causation. The decrease in the number of neural connections can be interpreted not only as becoming lazy but also as "cognitive efficiency optimization" (Cognitive Offloading). The decrease in brain activity may be due to outsourcing formatted and low - level cognitive tasks such as information retrieval to AI, thereby freeing up precious cognitive resources for higher - level strategic planning and critical research. The current experimental design cannot fully distinguish between these two completely different cognitive modes.

In short, a small - sample study over a short period of time cannot draw a universal conclusion. To derive the conclusion spread by the media that "AI leads to a decline in cognitive ability," more rigorous and long - term longitudinal studies are needed.

Meanwhile, some studies have concluded that "AI tools are helpful for human brain thinking activities," further supporting the view that "using AI does not lead to brain degeneration."

In February 2025, a study titled "The Impact of Generative AI on Critical Thinking: Self - Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers" by Microsoft Research and Carnegie Mellon University surveyed 319 knowledge workers and found that when users were confident in their professional abilities in a certain field, using AI actually stimulated more critical thinking.

A systematic review titled "Higher - order thinking skills - oriented problem - based learning interventions in mathematics: A systematic literature review" in the journal "Computers & Education" analyzed 69 relevant experimental studies from 2022 to 2024 and showed that using ChatGPT can improve students' critical thinking, problem - solving, and creative thinking abilities in higher - order thinking training. The study emphasized that AI - assisted problem - based learning (PBL) can help students integrate concepts and conduct logical reasoning more effectively.

What these empirical studies want to convey is that we need to go beyond the simple binary opposition of "AI is harmful or beneficial" and instead think about a more fundamental question:

How can we use AI to become better?

03 Embrace AI, Not Be Domesticated by It

Throughout history, every technological revolution has triggered fears of human ability decline. The ancient Greek philosopher Socrates worried that writing would damage human memory. The book "Amusing Ourselves to Death" warned that television would lead to the decline of linear logical thinking. Marshall McLuhan pointed out that computer systems might completely paralyze our central nervous system.

However, writing, television, and computers have not made us stupid; instead, they have improved our skills.

Writing has indeed changed the way we remember, but it has not eliminated our memory ability. Instead, it has created a more complex knowledge system and recorded precious human civilization. Although television has changed the information - receiving mode, it has cultivated new visual thinking abilities and enriched our understanding of the world. Computers have not paralyzed our nervous system; on the contrary, they have greatly expanded the boundaries of human cognition and turned the world into a global village.

The view spread by the media that "AI makes people stupid" often implies the misunderstanding of technological determinism and ignores human subjectivity in the use of technology.

It is undeniable that the brain is like a muscle: the more you use it, the stronger it becomes; if you don't use it, it will degenerate. When writing independently, the brain goes through a deep cognitive processing process such as "conceiving, organizing language, expressing, and revising." When using AI, one only makes the most superficial Prompt input without thinking about the logic and context behind the answer.

However, the impact of AI on cognition is not a one - way decline or enhancement but highly depends on the way of use and educational design. Just as some people use computers to learn courses from around the world while others become addicted to games and develop psychological problems, the key lies in how we use technology.

If an individual blindly, passively, and comprehensively relies on AI to complete all cognitive tasks, in the long run, their ability to independently solve complex problems, conduct in - depth logical thinking, produce original results, and have effective memory may decline substantially.

However, the impact of AI on thinking varies from person to person and from the way of use. Users who can actively use AI as an auxiliary tool rather than a complete substitute, maintain critical thinking during use, and consciously engage in cognitive exercise and in - depth participation may be less affected negatively and may even be able to use AI to improve their cognitive efficiency and creativity.

If we can adhere to critical thinking while using AI for brainstorming and maintain the habit of in - depth memory writing, AI will not make us stupid but will instead be a more intelligent search - engine assistant.

So, when we revisit the MIT research report, we will find that what it wants to express is:

Technology is neutral, and its impact depends on the way of application.

However, to prevent addiction, the next time you are about to ask AI a question, you might as well ask yourself: "If AI did not exist, how would I solve this problem?"

Perhaps, the answer lies in your own brain.

This article is from the WeChat official account "Brain Extreme Body" (ID: unity007), author: Coral. Republished by 36Kr with permission.