- Ravi Prakash
Imagine you have just typed out a long message and sent it. Two minutes later, someone asks, “Can you tell me what you just wrote?” and you cannot remember. It seems strange, doesn’t it? Yet this very question is now being debated seriously in the context of artificial intelligence. Is AI making us smarter? Is it boosting our productivity? Or are we gradually handing over all the thinking work we are supposed to do to AI?
Two recent studies from MIT have made this debate even more interesting. One examined how our brain works when we write using ChatGPT. The other looked at what happens when we use AI to detect fake news. Both studies raise the same uncomfortable question: when AI does more of the work, what happens to our cognitive skills?
Based on these studies, there is still no scientific proof that AI is damaging our brains. The first MIT study is a preprint, which means it is not yet final, peer-reviewed research, and its sample size is relatively small. So this is not a final verdict. It is only a warning signal.
The First Experiment: Brain Activity While Writing
MIT researchers divided participants into three groups. The first group had to write essays using a large language model (LLM) such as ChatGPT. The second group had to use a search engine. The third group could not use any external tool and had to write the essay with their own brain alone.
The researchers did not look only at how well the essays were written. They also used EEG technology, which allowed them to measure how different areas of the brain communicated with each other while the essay was being written. An interesting pattern emerged. The brain-only group showed the strongest and most widely distributed neural connectivity. The search engine group fell in the middle. The LLM group showed lower overall connectivity.
This must be understood very carefully. It does not mean that using ChatGPT damages the brain. What the researchers observed was that the more work the technology did in a task, the less measured brain engagement there was during that task. Put simply, the research suggests that cognitive engagement may be linked to how much of the work we do and how much AI does.
But the story does not end there. The researchers did not only look at brain activity. They also questioned participants about the essays they had written. Here, participants who had used the LLM struggled the most to accurately recall the content they had written. The essay was in their name, but when asked what they had written in it, they found it difficult to say.
Consider what that means. You write an answer and submit the paper. Then the examiner asks you to say what you wrote, and you do not remember. This is where the real question begins: did you actually learn something, or did you merely receive an output?
To explain this, the MIT researchers use a concept called “cognitive debt.” AI may have made your work easier today. But if we repeatedly hand over to AI the mental effort we are supposed to put into that work, could there be a long-term cognitive cost? That is the question.
A Big Catch: Sensational Headlines
There is a big catch here, and this is where some headlines turned very sensational. “ChatGPT is making you stupid.” “AI is destroying your brain.” Verdicts began to be handed down. Commenting in this way is not a scientific conclusion.
In the first study, there were 54 participants in the initial sessions, but only 18 by the fourth session. The paper had also not completed peer review, and the participants came from a relatively limited background. So based on this one study alone, we cannot say that AI is making everyone dull-witted.
However, the experiment does raise one important question: the more AI thinks for us, the less we may be thinking ourselves?
The Second Experiment: Can AI Catch Fake News?
Around the same time, the MIT researchers conducted another experiment. This time the topic was fake news. In this research, 67 participants were observed over four weeks. They were shown news headlines and images, and the question was simple: is this real news or fake news?
In some cases, participants were allowed to use an AI chatbot. In others, they had to decide on their own. At first, AI genuinely helped a great deal. With AI assistance, participants were 21 percent more accurate at identifying fake news.
This is a very important point, because the research is not saying that AI is always bad. Used properly, AI can improve our performance. It can help us analyze information, identify patterns, challenge our assumptions and increase accuracy.
Where the Problem Emerged: When AI Was Removed
So where did the problem arise? It arose when AI was taken away. What happens to our performance when AI is removed? After four weeks, participants were given new news items without AI, and the researchers noticed something interesting. Their independent ability to detect misinformation had dropped by 15 percentage points compared with their performance at the start of the study.
In other words, performance is better when AI is present and weaker when AI is removed. Researchers are discussing this as the “AI dependency paradox.”
This is a very important idea, because the problem may not be that AI gives wrong answers. The problem may be that we are gradually losing the ability to find answers without AI.
One in Five Developed Dependency
The researchers identified another interesting pattern. About one-fifth of the participants showed what was called a “dependency developer” pattern. At first, they investigated information themselves. But as time went on, they began to accept AI guidance more and more. This is where the real warning lies: we may lose the habit of thinking without AI.
Take GPS. GPS is extremely useful. But if you use it for every little journey, you may stop observing the roads and directions around you. GPS technology has not damaged your brain. You have simply practiced one particular skill less. The same principle may apply to AI.
Not the First Time Technology Has Changed Our Thinking
Here the story gets bigger, because humans have always been handing over part of their mental tasks to technology. The calculator arrived, and manual calculation declined. Search engines arrived, and the need to remember information declined. GPS arrived, and remembering routes declined. Now AI has arrived, and it can handle writing, summarizing, translation, research, analysis, coding, planning and many other tasks.
Every time a new technology appears, the same question is asked: will humans become less intelligent? But history gives a complicated answer. Technology took some skills away from us. At the same time, it also gave us new capabilities.
Did the calculator destroy our math? When it arrived, similar fears were raised. It was said that children would no longer be able to calculate and that math skills would decline. But mathematics did not disappear because of the calculator. Technology changed the work we do. Instead of spending hours on repetitive calculations, we were able to focus more on complicated problems. The same thing happened with the computer, and with the internet.
That is why the real question about AI may not be “Will AI reduce our intelligence?” The real question may be “Which mental skills are we ready to outsource to AI?” Not all skills are the same. You can get AI to do multiplication. You can get AI to write an entire school essay. You can get AI to write a news story. But letting AI make the important decisions in your life is a completely different level.
The Real Problem: Cognitive Offloading
Scientists use a term for this: cognitive offloading. Put simply, it means handing over part of the mental task our brain is supposed to do to a tool outside us. Noting down a phone number instead of memorizing it, doing a calculation on a calculator, or checking a route on GPS are all forms of cognitive offloading.
We cannot say this is definitely wrong. Many times, external tools improve our performance. But what if we outsource a skill completely and, in the end, cannot perform it ourselves? That is where the real problem begins. The use of AI is now asking us exactly this question.
Should We Stop Using AI?
Perhaps not. Even the MIT research does not say that. In fact, the second study itself showed that when AI is used properly, fake news detection performance improved significantly. The real issue is how we are using AI.
One approach is to say, “Write my entire essay.” Another is to say, “This is my essay. Tell me the weaknesses in my argument.” There is a huge difference between the two. One gives you an answer, while the other makes you think.
Likewise, asking AI, “Tell me whether this news is fake or real,” is one thing. Asking, “What questions should I ask to find out whether this news is fake or real?” is another. In the first approach, AI gives you the answer. In the second, you have to find the answer yourself.
AI: Copilot or Substitute?
Imagine two students. Student A gave the entire assignment to AI. AI wrote it, he copied it and submitted it. Task complete. But what did he learn?
Student B first wrote a draft himself. Then he asked AI, “What are the weaknesses in my argument?” Then, “What is the strongest counterargument to this?” Then, “What am I missing?” Here, AI is not thinking on the student’s behalf. It is challenging the student’s thinking. This is the most important difference.
MIT researchers also believe that an AI system that asks questions, a Socratic approach, rather than giving users direct answers, can better support independent reasoning.
So, Is AI Making Us Dumb?
The scientific answer is that we do not currently know. Based on these studies, we cannot say that AI is permanently damaging our brains. But the research signals suggest that the way we use AI may affect cognitive engagement, memory, learning and independent performance.
That is why this debate is important. The future is not humans versus AI, because AI is already with us. The real question is what kind of humans we are going to become while using it.
The Real Danger
Perhaps the real danger is not AI itself. Perhaps the biggest danger is not that AI becomes very intelligent. The real danger is that it becomes very comfortable for us to let AI think for us.
If AI writes everything, our writing skills may decline. If AI researches everything, our research ability may decline. If AI verifies everything, our questioning may decline. If AI makes every decision, our judgment may weaken too. Then technology changes from a tool into a substitute.
The Responsibility Is Ours
Will AI destroy our brain? Based on the evidence so far, we cannot say that. But can AI change how our brain engages with difficult tasks? Yes, and there is research evidence for it. That is why the responsibility is ours.
There is no need to stop using AI, but we need to change the way we use it. We should think first, then ask AI. We should challenge the answer and verify the information. And sometimes, we should switch AI off and do the same task ourselves. Just as a muscle grows weak if it is not exercised, our thinking ability stays strong only through practice.
So it will not be those who use AI the most who succeed in the future. The advantage will belong to those who know when to use AI and when to set it aside and think for themselves. AI should not become a replacement for our brain. Our brain needs a sparring partner. And that, perhaps, is the future in which humans and AI think together, and think better.