Is AI Changing the Way We Think?
Imagine waking up tomorrow with an extraordinary new ability.
Whenever you encounter a difficult problem, you can get an answer almost instantly. Struggling to write an important email? The right words appear in seconds. Trying to understand a complicated subject? You receive a clear explanation without spending hours searching for one. Faced with a problem you don't know how to solve? You have a solution before you've had much time to think.
At first, it feels as though you've become more intelligent. You accomplish more, work faster and find yourself capable of things that once seemed difficult.
Now imagine using this ability every day for an entire year. It becomes such a natural part of your routine that you barely notice when you're relying on it.
Then, one morning, it disappears.
You sit down to complete a task you've done dozens of times before, but this time, there's no immediate answer waiting for you. You have to rely entirely on your own knowledge, memory and reasoning.
How much of what you've accomplished over the past year could you still do independently?
Were you becoming more intelligent, or simply becoming better at accessing intelligence outside your own mind?
This scenario may sound imaginary, but the technology behind it is already becoming part of everyday life. Artificial intelligence can write, summarize research, solve problems and generate ideas in seconds. As these systems become more capable, much of the public debate has focused on whether AI could eventually become intelligent enough to replace humans.
But perhaps we should also be asking a different question: What happens to human intelligence when we begin relying on machines to do our thinking?
Is Artificial Intelligence Really Like Human Intelligence?
In his 2026 book Le grand mirage: Pourquoi l'intelligence artificielle ne remplacera jamais l'humain, Canadian neurosurgeon and researcher Dr. David Fortin challenges the idea that increasingly powerful AI is necessarily becoming more like the human mind.
Consider the extraordinary ability from our opening scenario. It can provide convincing explanations, construct arguments and solve complex problems. From the outside, it may appear remarkably intelligent. But does producing an intelligent answer necessarily mean understanding it in the same way a human does?
Many modern AI systems rely on artificial neural networks, computational structures inspired loosely by the connections between neurons in the human brain. These systems learn statistical patterns from data and use those patterns to generate responses, recognize images or make predictions.
Despite the name, artificial neural networks are not miniature biological brains. The human brain contains billions of living neurons and other cells communicating through electrical and chemical signals. It continuously integrates information from the body, senses, memories and environment.
One way to understand the distinction is to think about flight. Airplanes and birds can both fly, but an airplane does not need feathers, muscles or a biological nervous system to accomplish what a bird does. Similarly, an AI system can perform tasks associated with human intelligence without necessarily reproducing the biological processes underlying human thought.
Fortin argues that these differences are fundamental, particularly when it comes to consciousness: our capacity to have subjective experiences, such as feeling pain, experiencing emotions or being aware of our surroundings. An AI system might describe sadness convincingly, but does that mean it actually feels sad? It might explain how a person makes a difficult decision, but does that mean it experiences uncertainty or understands the consequences in the way a person does?
Fortin maintains that genuine human intelligence and consciousness are inseparable from living biological systems. However, whether consciousness requires biology remains an unresolved scientific and philosophical question. Researchers still do not fully understand how conscious experience arises, making definitive predictions about artificial consciousness difficult.
Regardless of whether AI will ever think like a human, it is already capable of performing many tasks that previously required human thought. This brings us back to our imagined ability and a more immediate question: what happens when we become accustomed to relying on it?
When Thinking Becomes Optional
Think back to that imagined year of effortless productivity. At first, you use your new ability only when you're stuck, perhaps to understand a difficult concept or organize your thoughts before writing something important. But as time passes, you begin using it for smaller tasks. Why spend twenty minutes drafting an email when it can be written in seconds? Why struggle through a complicated explanation when you can receive a summary immediately?
You aren't necessarily becoming less capable. You're simply choosing the easier, faster option. Psychologists have a term for transferring mental work to an external tool: cognitive offloading.
We do it constantly. When you save a phone number instead of memorizing it, use a calculator rather than doing arithmetic in your head, or write a reminder so you don't forget an appointment, you're relying on something outside your brain to help perform a cognitive task.
Imagine using GPS to navigate an unfamiliar city. It helps you reach your destination without memorizing every street or planning each turn. That's useful. But arriving successfully doesn't necessarily mean you could find the same destination again without assistance. The important distinction is between completing a task successfully and developing the ability to complete it independently.
Cognitive offloading is not inherently harmful. In fact, research suggests it can be beneficial. In a 2025 meta-analysis published in Memory & Cognition, researchers Lois K. Burnett and Lauren L. Richmond examined how external memory aids affect performance. Their findings showed that cognitive offloading can improve performance on memory-based tasks and reduce differences between individuals.
External tools can therefore expand what we accomplish. They allow us to work around certain limitations and devote attention to other activities. But generative AI introduces an interesting possibility. Unlike a reminder or a calculator, it can produce entire explanations, arguments and solutions. It can assist not only with remembering information or performing calculations, but also with the reasoning involved in completing a task.
Consider what happens when you learn something difficult. At first, understanding a new concept may require several attempts. You read an explanation, realize you don't understand part of it, return to the beginning and try to connect it with something you already know. Eventually, the idea begins to make sense.
That process involves mental effort, and practice can help strengthen our ability to use what we've learned. The brain can change in response to experience, a property known as neuroplasticity. Think about learning a new language. At first, forming a simple sentence might require considerable concentration. With repeated practice, recalling words and constructing sentences can become easier as learning changes how the brain processes the information.
This doesn't mean that every difficult task is beneficial or that making something easier prevents learning. But it does raise a question about our imagined year of AI assistance: if a tool regularly performs the reasoning for us, what opportunities do we still have to practice that reasoning ourselves?
What Happens When the Assistance Disappears?
Now return to the morning when your extraordinary ability is suddenly gone.
You open your computer and begin writing. For months, you've relied on AI to organize your ideas and turn them into polished paragraphs. Without it, you may discover that some tasks still feel familiar, while others require more effort than expected.
But what would science actually predict?
Researchers are beginning to investigate how generative AI affects cognitive engagement, learning and independent performance. So far, the evidence is more complicated than the idea that AI simply makes us smarter or less intelligent.
In 2025, researchers associated with the MIT Media Lab conducted an experiment involving 54 participants completing essay-writing tasks. Some used an AI language model, others used a search engine, and a third group wrote without external assistance. The researchers measured electrical activity in the brain using electroencephalography, or EEG, while participants worked.
Across the initial sessions, participants writing without external assistance generally demonstrated stronger and more distributed patterns of measured brain connectivity than those using AI. Participants in the AI-assisted group also had more difficulty recalling parts of their own essays.
Imagine two people producing equally polished essays. One develops an argument independently, while the other receives substantial help from AI. The final essays might look similar, but the mental processes involved in creating them could be quite different. The MIT findings suggest that AI assistance may influence cognitive engagement during writing. However, the study was relatively small, focused on a particular task and was released as a preprint. It does not establish that AI causes permanent cognitive decline or that using it for other activities would produce the same results.
Another 2025 study, conducted by researchers from Microsoft Research and Carnegie Mellon University, surveyed 319 knowledge workers about their use of generative AI. The researchers found that people who expressed greater confidence in AI tended to report engaging in less critical thinking during AI-assisted tasks. Yet participants also described using their judgment to evaluate, verify and integrate AI-generated information.
This suggests that AI may change where our mental effort goes. Instead of producing every answer ourselves, we may spend more time deciding whether an answer is reliable or appropriate. Because the study relied on self-reported experiences, it cannot establish that AI directly caused these differences. Still, it highlights an important possibility: using AI does not necessarily eliminate thinking, but it can change the kind of thinking we perform.
Research on education offers another perspective. In a 2026 longitudinal study, Chrysanthi Melanou, Maik Beege and Martin Kimmig followed 87 university students over a semester, comparing AI-assisted learning approaches with a group that did not use AI.
All three groups improved their knowledge, but the researchers did not find an additional overall learning advantage from AI integration. Critical thinking remained relatively stable, while more reflective engagement with AI was positively associated with critical thinking. The study was relatively small and limited to one educational context, so its results cannot be generalized to every student or learning environment. Nevertheless, it suggests that simply having access to AI does not guarantee better learning.
Taken together, these findings cannot tell us exactly what would happen after a year of relying heavily on AI. They do, however, suggest that we should distinguish between the quality of the work we produce with assistance and the knowledge or abilities we develop along the way.
Perhaps the most important part of our imagined scenario, then, isn't the moment the assistance disappears. It's what we were doing while we had access to it.
What If We Used the Same Ability Differently?
Imagine starting that year again.
You still have access to the same extraordinary technology. It can answer questions, explain difficult concepts and help you accomplish tasks faster than before. But this time, you decide to use it differently.
When you encounter a difficult problem, you attempt it yourself before asking for help. Instead of requesting a completed answer, you ask the AI to identify mistakes in your reasoning. When it explains something unfamiliar, you try to describe the idea in your own words and test whether you can apply it to a new situation.
The technology hasn't changed. What has changed is the role it plays in your thinking.
This distinction is especially important in education, but it applies far beyond the classroom. A student might use AI to practice explaining scientific concepts, a professional might use it to challenge an argument, and a researcher might use it to explore alternative interpretations of evidence. In each case, AI can support intellectual work without necessarily replacing the person's responsibility to understand, evaluate and make decisions.
Of course, asking AI to challenge our thinking doesn't guarantee that we'll learn more. AI systems can make mistakes, offer misleading explanations and generate convincing arguments that are not supported by evidence. Their usefulness still depends partly on our ability to question what they produce.
Researchers are continuing to investigate which forms of AI assistance improve learning and which may encourage excessive dependence. Long-term studies will be especially important for understanding whether changes in everyday cognitive habits lead to lasting changes in independent abilities. For now, the evidence does not justify treating AI as either an inevitable threat to human intelligence or an automatic enhancement of it.
Return, one final time, to the extraordinary ability from the beginning. You can still obtain answers in seconds and accomplish more than you could alone. But now you recognize something that may not have been obvious when you first discovered it: having access to an answer is not the same as understanding it.
Perhaps some of the work you once struggled through can be delegated without meaningful loss. Perhaps other tasks are worth practicing precisely because the process of working through them helps develop abilities you want to retain. Science has not yet established where all those boundaries lie.
Fortin's argument challenges the assumption that increasingly capable AI must eventually become equivalent to human intelligence. Whether his conclusions about consciousness prove correct remains uncertain. But his concern about our relationship with technology raises a question we don't have to wait for artificial consciousness to investigate.
AI does not need to think like us to change the way we think. It only needs to become useful enough that we stop noticing which parts of our thinking we've begun handing over.
The Nexus Question
As AI becomes better at thinking for us, how much of our own thinking can we afford to give away?
Sources & Further Reading
David Fortin — Le grand mirage: Pourquoi l'intelligence artificielle ne remplacera jamais l'humain (2026)
A neuroscientific perspective on the differences between human and artificial intelligence, consciousness and technological dependence.
Éditions La Presse — Le grand mirage
Lois K. Burnett & Lauren L. Richmond — Meta-analytic Investigations of the Effect of Cognitive Offloading on Memory-Based Task Performance and Interindividual Variability (2025)
A meta-analysis examining the benefits and limitations of using external memory aids.
Nataliya Kosmyna et al. — Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Tasks (2025, preprint)
An experimental investigation of brain connectivity, recall and cognitive engagement during AI-assisted writing.
MIT Media Lab — Your Brain on ChatGPT
Hao-Ping Lee et al. — The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers (2025)
A survey exploring the relationship between AI confidence, cognitive effort and critical-thinking practices.
Microsoft Research — Full Study
Chrysanthi Melanou, Maik Beege & Martin Kimmig — Generative AI and Learning Dynamics in Higher Education: A Longitudinal Empirical Study (2026)
A longitudinal study examining knowledge development, critical thinking and different approaches to AI-assisted learning.
