When Assistant Professor Bob Reuter began studying cognitive psychology in the 1990s, neural networks were a curiosity. Now, as Head of the Institute for Teaching and Learning at the University of Luxembourg, he explores how such systems reshape how we learn, teach and think.
For the last twenty years, Reuter has examined how digital technologies reshape humanity’s relationship with knowledge and power. “The book has already changed how we conceive of being educated, what it means to be literate and intelligent. The internet and user-generated content have reshaped that again, and generative AI is most probably disturbing all this once more.”
That fascination with how minds, both human and artificial, learn began much earlier in his career. Researchers at the time used primitive simulations of the brain’s cognitive processes, models which revealed striking parallels with human development. “Take how children say ‘I goed’ instead of ‘I went’. Adults don’t speak like that, so imitation doesn’t explain it,” Reuter recalls. But an artificial neural network exposed to English rules did exactly the same thing. It applied the regular -ed rule everywhere. It was a revelation that helped researchers model learning itself.
Three decades on, the Assistant Professor views today’s large language models (LLMs), from ChatGPT to Gemini, as something else entirely. “It’s important to see what we do with them, and then how they become hammers and scissors of the mind,” he says.
There’s growing concern that heavy reliance on generative AI tools could indirectly affect short-term memory and related cognitive functions. To what extent would you say that the memory aspect is an alarming part of this disruption?
Bob: I wouldn’t call it alarming. It’s part of who we are as a species to have an extended mind. Like the beaver and its dam, you cannot understand the animal without understanding the tool it builds. For humans, being able to extend our memory into physical objects has been part of what it means to be human. Writing, reading, storing on digital supports, that’s all part of being an extended cognitive species.
So you don’t agree that memory is being impacted?
Bob: Already 3,000 years ago, people were worried that writing would destroy memory, and I think writing has done just the contrary. We don’t store information only in the brain; we store it in other humans, in songs, in things we write down, in what we keep on digital devices. Our collective and individual memory has not decreased, it has increased.
Some people worry that relying on tools like translators makes us forget how to write or think in other languages. What do you say to that?
Bob: I would expect that, because we are what we do. If you don’t do the translation with your brain, the capacity disappears progressively. But is it that bad? My parents used paper maps to plan holidays. Now we use navigation tools. Of course it’s a pity from a nostalgic point of view, but as long as you have access to the satnav, you can still do the task.
You’ve said that what we do shapes who we are. How does that apply to AI use?
Bob: If you do certain cognitive tasks a lot, your brain will adapt. The brain is plastic, it’s the protocol of its use. But I would be cautious about claims of quick or long-term effects. Learning from experience is a slow process, in that the brain as evolved to be relatively resilient and not to let itself be reshaped by individual extraordinary experiences, except in the case of trauma. But, in general, learning and development is slower to protect the brain from going into dead ends.
People often say AI will free us to do “other things”. Do you agree?
Bob: The problem is that this rarely happens. We’ve seen the myth that using digital tools in education will free teachers’ time. No, it will not, because you’ll use that time to be more productive. The expectation will be that now you can do more.
What can people do to protect and strengthen their cognitive abilities?
Bob: If you use tools to help you, you still want to develop your ability to think critically. That’s even more important now. It’s so easy to fall for stuff that sounds plausible. Read from different sources. Read the primary articles. And be cautious when you talk to a chatbot. They are designed to be agreeable. If it agrees with you, there’s probably something to be critical about. They mimic empathy and put you in the driver’s seat, but they rarely say what a good teacher should: “No, you’re wrong. Think again.”
So use them differently?
Bob: Yes. I recommend using Socratic or tutor modes. These can act as your critical friend, asking, “How do you know that?” That’s something you can also do without an LLM.
What about the effects on empathy or social cognition?
Bob: My first response would be: how could we already know? The phenomenon is new, and there are probably not enough studies. If people feel lonely and turn to chatbots, that’s more a symptom of our society than the cause of all evil. In the 1990s, video games were blamed for spoiling youth. But humans were violent long before writing.
Could chatbots ever have a positive emotional role?
Bob: If people can reduce loneliness or depressive symptoms with the help of a benevolent chatbot, then why not? Someone could imagine a chatbot designed for people who don’t dare to interact with others, to prepare them for the real world. That could be very valuable.
What worries you most about AI in education?
Bob: The elephant in the room is that the dangers are not in the tools themselves, but in the humans and systems within which they are embedded. We’re allowing private companies, whose main value is shareholder value, to run education. Learners use chatbots because they adapt to their language, which is fantastic. But we’re delegating access to knowledge to entities driven by profit. That worries me much more than many other things.
This article was published in the Silicon Luxembourg magazine.
