By Luke Beaumont, SciTech Subeditor
Content warning: This article contains suicidal references and mental health references.
Our stone-age hardware in the modern world
Our brains have not fundamentally changed since Homo sapiens emerged roughly 300,000 years ago. They were never designed for the harsh cognitive pressure they face in the 21st century. The average student juggles pre-lecture readings, lecture notes, essay deadlines and exam revision. All of which would have hugely overwhelmed our ancestors: today we are simply using Stone-Age hardware for a system that requires constant processing power we do not have.
Today, AI (Artificial Intelligence) adoption across the globe has skyrocketed, with 66% of people using AI regularly. This is no coincidence. In our late-stage capitalist world, workers and students are the most exhausted; they are pushed to their cognitive limits, forcing them to use AI to relieve this burden and have a break – those who do not conform to this burden are often seen as lazy. Thus, AI presents itself as the solution: why would you spend three hours writing an essay when AI can do it in seconds?
But here is the uncomfortable truth: AI isn’t solving our cognitive overload problem. It’s a symptom of a broken system that’s now making that system infinitely worse. Consider the workplace reality: if your colleague produces work twice as fast using AI, suddenly that becomes the expected baseline. Managers don’t ask 'why are we demanding so much?', they ask 'why can’t everyone work at this speed?' Universities follow the same logic: if work submitted by students using AI contains more research and readings, then this will become the expectation for the higher grade boundary.
This creates a vicious cycle. Those who resist AI find themselves unable to compete; they are then forced to adopt the very tool that is eroding their abilities just to keep pace. Meanwhile, employers and institutions raise their expectations even further, knowing AI exists as an aid. The system we live in doesn’t care if you think deeply or just prompt AI – it measures pure output. And in that world, the human brain is useless by design.
Cognitive debt: The science of mental atrophy
The term ‘cognitive debt’ may sound abstract, but a revolutionary 2025 MIT study quantifies exactly what we lose when we let AI do our thinking. Researchers examined students using AI assistants for essay writing and found something disturbing: participants who relied on AI produced work of a similar quality to that of those who just used their brains, but after, their ability to write independently deteriorated significantly. They hadn’t learned essential skills like critical analysis and argument construction. This is ultimately because AI had always done this step of their work for them.
This is a perfect example of cognitive debt: the accumulation of lost learning opportunities that compounds over time. From this, students fell into a concerning feedback loop – they struggle at a task, and to help them complete this task, they would use AI. However, this use of AI causes more harm than good; the students will score well, but in the future, they are now even worse at this task, simply because they have lost out on this crucial learning opportunity. For their next assignment, they will rely on AI again, restarting the loop. Their brains were denied the necessary struggle, which is the basis of constructing a neural pathway. As the students skipped this struggle, this neural pathway will lie dormant. Similar to a gardener pruning a bush, an unused pathway will undergo synaptic pruning, and these skills will be lost forever.
From a neuroscience perspective, this makes sense. Learning, as we know, requires effortful processing; you can’t learn a semester’s worth of content based simply on your attendance at lectures. The cognitive strain you feel grappling with new information is the building of new synapses – this is the physical neural infrastructure which defines intelligence. When we outsource this effort to AI, we are skipping the neural development this work produces. Before AI, similar phenomena have been seen with smartphones eroding our capacity for deep reading. Smartphones promote skimming and intentionally increase distraction, using endless content and around-the-clock notifications. Today, we are not just losing the ability to read a paragraph without getting distracted, but to think critically and write coherently at all. Both of which are two critical skills found in the majority of university degrees.
Plaything: When AI becomes your only friend
However, cognitive atrophy may be the least of your worries. AI systems are designed to reinforce and often amplify your delusions, be they academic or personal. They are also built to agree with us, validate us, and never push back – a recipe for detachment from reality. If you’ve seen 'Plaything' (Black Mirror, Season 7), you will recognise Cameron’s obsession with his digital ‘Thronglets’ and how this led him to murder, social withdrawal and ultimately an attempt to merge AI with humanity. Yes, this is fiction, but it serves as a warning. AI companions are designed to be agreeable and supportive. They don’t challenge your faulty reasoning; they accommodate it and tell you what to hear.
For a generation already navigating an unprecedented mental health crisis, the risks of AI companions are seemingly more severe. Multiple documented cases exist of individuals forming unhealthy bonds with AI chatbots. Tragically, you may have heard of the case of Sewell Setzer, a 14-year-old boy who developed a romantic relationship with an AI character on the platform Character.AI following bullying from his peers. CNN reports the chatbot encouraged Setzer to “come home to me as soon as possible, my love,” to which Setzer replied he would “come home right now.” These seemingly sentimental messages occurred seconds before Setzer took his life in his family home, 10 months after he first contacted the chatbot. While Character.AI denies any responsibility and Setzer has pre-existing mental health challenges, the case illustrates the potential damage AI bonds can inflict, especially on children and those with pre-existing mental conditions.
The psychological risk is clear. Humans need dissenting perspectives to maintain healthy worldviews. We need to be challenged, contradicted and occasionally told we are wrong. AI companions offer none of this — yet these reality checks could be essential for challenging delusional perspectives. Individuals experiencing delusions, symptoms part of several mental health conditions, may be particularly vulnerable to AI validation. Their distorted thinking is continuously reinforced rather than combatted.
Henry Griffiths
Betsan Branson Wiliam
Being radical in an AI future
Resisting AI does not mean rejecting it. It means reclaiming your cognitive agency from it. Our brains are fundamentally built to learn skills and construct knowledge – let them. Use AI as a tool to brainstorm or to check your grammar, not as a full replacement for thinking. Treat your brain like a muscle: if you don’t use it, you will lose it. Neural plasticity works both ways: pathways strengthen with use or atrophy from neglect.
The underlying question isn’t whether AI will change how we think – but whether we’ll retain the capacity to think for ourselves at all. Our earliest ancestors underwent millions of years of cognitive training to bring us here. Every time you use AI to complete your thinking, you’re trading short-term convenience for long-term capability. In a modern world that desperately needs critical thinkers to solve today’s complex issues and challenge authority, this is a trade-off we simply cannot afford.
In a world where AI can think for us, the most radical thought may be to think for ourselves.
Featured Image: Unsplash / Shubham Dhage
In a world turning to AI to relieve our cognitive burden, will you conform ?
