When the first navigation systems were introduced into our cars in the early 1990s, I remember reacting with a mixture of scepticism and relief. I was sceptical because, at the time, I wasn’t yet convinced that this new technology could truly prevent me from ending up in the wrong place. Yet I also felt relieved, because part of me hoped that, at last, I might finally stop ending up in the wrong place!

More than thirty years later, I rarely get lost while driving. At the same time, I’ve become almost incapable of remembering a single route I’ve taken from A to B.

Back then, I used maps to figure out the best way to get from one place to another, relying on landmarks and my own sense of direction. Quite often, that meant heading roughly where I thought my destination was and then exploring the surrounding area,  on foot or by driving around, until I found it. It took longer, but once I’d found the place, I could return to it again and again without a moment’s hesitation.

Look at me now. On the rare occasions when there is no signal and the GPS stops working I feel a genuine sense of panic. My sense of direction struggles to kick in, and in unfamiliar places I hardly trust it anymore. Second guessing myself in favour of what my GPS tells me has become a habit. I’ve grown so accustomed to technology guiding me that I barely pay attention to where I’m actually going. As a result, I’ve become increasingly dependent on it and, I feel, less and less capable of finding my way by relying solely on my own sense of direction.

Could this be a metaphor for what’s happening to me more generally as I use Artificial Intelligence for an ever-growing variety of tasks and across more and more domains?

In their study on “The effects of over-reliance on AI dialogue systems on students’ cognitive abilities” (2024), Zhai et al. argue that heavy reliance on AI can reduce:

• critical thinking
• problem-solving
• creativity
• memory retention

Research led by MIT appears to support the same conclusions, as do several other studies. But even without that evidence, I only have to rely on my own observations to reach a similar conclusion: as far as I’m concerned, I’ve become intellectually lazier. Not in every respect, but I find myself taking mental shortcuts whenever artificial intelligence offers an easier and faster alternative.

For example, in the past, when writing an important text in a language that wasn’t my own, I would ask a trusted native speaker to review specific passages. I wouldn’t just ask them to correct my writing; I would also ask them to explain why one expression worked better than another. My goal was not only to produce a polished, professional-looking text, but also to learn from my mistakes so that I could write better the next time. Today, I rely heavily on AI to perform the same task. There’s nothing wrong with that. After all, the ideas are still mine, the arguments are mine, and the text is mine. The difference is that, whereas I initially asked AI to improve only the passages I found problematic, I now routinely ask it to review the entire text and make it “more fluid.”

But what was wrong with the way I had written it in the first place? Had my writing really been so poor before the advent of artificial intelligence? I don’t think so. Yet my brain seems to find that question irrelevant.

The same applies to writing summaries and adapting the same text for different audiences. In the past, I would consciously put into practice what my teachers had taught me: the invaluable skills of synthesis and adaptation. I would take the time to understand my audience, who they were, what they needed, and how they were most likely to engage with my message. I could spend hours refining a single document if that was what it took to distil its essence into just half a page.

Not anymore. It has simply become too easy, too fast, and too effective to ask AI to do it instead. Why should I bother? I’m saving a considerable amount of time. I can publish more often, adapt the same content for different platforms, and spread my message much more widely. I should be delighted. I’m using my time more efficiently than ever.

The problem, as I see it, is that while I am busy using AI, I also have the unsettling feeling that I am slowly losing some of my own competence. The brain is not a muscle but it works in much the same way: what you train becomes stronger, leaner, faster, and you are more likely to be able to rely on it when you most need it. Any muscle you stop training (and using) becomes less reactive and weaker. I don’t need a huge amount of statistics to confirm that translating a document without AI is now more demanding than it used to be, or that writing a summary of an article requires more mental effort than it once did. I could list many other examples of tasks that AI now performs for me. This change is not (yet) due to the natural aging process. It seems closely linked to the fact that, because AI’s cognitive capabilities are constantly within reach, I have fewer reasons to exercise my own. AI doesn’t just save time; it removes the practice. And without practice, even well-developed abilities begin to fade. The less I use them, the less capable I become of relying on them when I truly need them.

I don’t want to say that this is all bad. While we are losing some skills and competences, we are also acquiring new ones that make us better adapted to the times we are living in. And yes, that is exciting, new, and full of potential. However, I want to make sure that, among all the skills and competences that I’m losing along the way, critical thinking is not one of them.

According to the Neuroleadership Institute, critical thinking is the process of:

1. Evaluating information
2. Questioning assumptions
3. Analysing evidence and
4. Making reasoned judgements

All in the service of reaching valid conclusions. 

It is now more important than ever that we, as human beings, continue to exercise this essential ability. We need to guide, refine, and assess the reliability of AI and to question how much trust we place in its answers and suggestions. We need to be vigilant to recognise and challenge the biases inherent in the data on which AI is trained. Most importantly, to avoid the risk of becoming a little less competent every day, we need to resist the temptation to always take the easy route of saying, “Let me ask AI.” Instead, we should regularly challenge ourselves to do things the old-fashioned way: by using our own brains.

Sources:

Chunpeng Zhai, Santoso Wibowo1 and Lily D. Li, “The effects of over-reliance on AI dialogue systems on students’ cognitive abilities: a systematic review”, Smart Learning Environments, 2024.

Massachusetts Institute of Technology (MIT), “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task”,  June 10, 2025, https://www.media.mit.edu/publications/your-brain-on-chatgpt/

Fakhroddin Noorbehbahani and Kiemute Oyibo “AI-overdependence and human cognitive decline: Hazards, evidence, and mitigation strategies”, Computers in Human Behaviour Report, Vol 22, May 2026.

Neuroleadership Institute, “AI Can Give You Answers, But It Can’t Think for You: The Neuroscience of Critical Thinking”, webinar, 11 March 2026.