There are a few tasks where I find AI tools like ChatGPT or Claude indispensable. AI can be very helpful in finding Swiss (Tilsiter) cheese holes in my logic and argumentation. My favourite new sport (and form of procrastination) is to take the feedback from one tool and feed it to another, asking what the other one missed or got wrong. For me, this is the first level of feedback that sometimes makes sense and saves time for me and my supervisor. Some of the AI feedback will be spot on; some will make sense only from a different perspective, showing me that I didn’t make myself clear enough. Sometimes it forces me to think in a different direction and pushes me toward a new idea.

Then there comes the topic of text polishing. Even if MS Word can check my spelling and spot obvious mistakes, when it comes to more complex grammar issues, I ask for help. So, summing up – I use AI as a first-level support in terms of correction and sparring, as well as for tasks that are not taxing or intellectually defining.

Now let’s flip that coin – what is a no-go and why? For me, the red line is where it’s no longer about feedback and challenging my thinking, but about taking the creative work away from me. Some of the reasons are obvious, and the first one is ethics. Another reason lies in the very nature of intellectual work: I find it intrinsically motivating, and I like that shot of (expensive) dopamine when I manage to formulate an idea, when I have my little private a-ha! moment. Overcoming this type of challenge is incredibly rewarding in itself. Even if you find it debatable or call me a masochist, I think we can agree that brain work helps us become more clever, smarter, and intellectually stronger. All in all, it provides a gym for that little hamster in our heads to keep its fur sleek, maintain its muscle mass, and live longer – because it’s the very activity that keeps it going.

Even if certain things are done to achieve tangible goals – be it the next promotion or a pay raise – there’s an expectation that by gaining experience we’re becoming better thinkers and problem solvers. And of course, we hope that even if Alzheimer’s does pay us a visit one day, we can at least postpone that gruesome moment by staying intellectually fit. Hence, it would be a darn bad idea to give this experience and this ammo away. For here comes the crux: by offloading the challenge, we might be solving a short-term problem or, to be more precise, jumping the queue (submitting a concept on time), but we’re creating a much bigger problem for our future selves – the ones who might not even remember our name in 20 or 30 years. That, and making Idiocracy less of a dystopian satire and more of a reality. This is why I always ask not to do a complete rewrite for me, but to give me feedback.

To stress the point: there is a captivating (albeit still nascent) research field studying how AI tools affect memory, cognition, and learning, often framed through the lens of my favourite new term – cognitive offloading. One recent study, with a telling title, “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task,” illustrates this well. As the name suggests, participants were asked to write an essay. They were divided into three groups:

• LLM / AI group – wrote essays with the assistance of ChatGPT or a similar model
• Search engine / web group – used Internet search for support
• Brain-only group – wrote essays without any external aid, relying solely on their own memory and knowledge.

Unsurprisingly, the groups’ ability to recall their own output varied dramatically. The authors describe a kind of “cognitive debt” that accumulates when one offloads too much to AI – over time, reliance may weaken internal cognitive processes.

So for me, when interacting with ChatGPT, Claude or the like, it boils down to a simple question: will this prompt make me think harder and deeper, or will this prompt take my thinking away?