Source: The Conversation – UK
Deep in Book VII of Plato’s Republic, Socrates describes prisoners chained inside a cave, mistaking shadows cast on a wall by firelight for reality itself. They name the shadows, debate them and develop expertise about them.
The prisoners are completely, sincerely wrong, and they have no idea. The cave isn’t a place of stupidity, it’s a place of convincing, well-organised illusion. But Plato’s real interest wasn’t the cave, it was in the periagoge – a Greek word meaning the turning of the soul away from shadows and toward the light.
For Plato, this was education itself: not the filling of an empty vessel with facts, but a fundamental reorientation of how a person relates to truth and how they come to know that truth. The shadows persist but today they aren’t cast by firelight, they are generated by machines.
Large language models (LLMs), image making and AI-powered search produce outputs that are fluent, confident and immediate. But here’s the crucial difference from Plato’s original problem, his shadows were at least connected to something real.
What AI produces is different in that a language model has no built-in commitment to truth, only a statistical relationship to an enormous quantity of text. When it tells you something, it isn’t reporting, it’s composing.
The outputs can be correct. But they can also be wrong in ways that are structurally indistinguishable from being right. The shadow no longer flutters on a cave wall, it speaks now, and sometimes it speaks beautifully.
This is why periagoge – turning towards the light – matters more now than ever and why AI threatens it so quietly. Knowledge isn’t merely true belief, it’s true belief held for the right reasons, connected to the world through justification, evidence and process.
AI disrupts this at the root. It is useful precisely because it decouples output quality from the slow, demanding work of verification. You don’t need to consult a primary source, triangulate between perspectives, or sit with the discomfort of not yet knowing.
Bypassing learning GenAI poses many problems for learning. When an AI hands us an answer, we risk bypassing the process through which learning happens. We’ve received a product that looks like knowledge from the outside but is hollow at its core, it’s a shadow that convinces us of something we haven’t actually understood.
GenAI doesn’t just help us think. It thinks instead of us. And there’s growing evidence it’s making us measurably worse at doing it ourselves. The philosopher Miranda Fricker identified a harm she called “epistemic injustice”, the wrong done when someone is denied the tools to make sense of their own experience.
Writing in 2007, she couldn’t have imagined the form that harm might take two decades later. What we risk now is something adjacent but distinct: epistemic atrophy. Not the theft of knowledge, but the slow erosion of our willingness and capacity to undertake the more demanding work of understanding what is real.
In other words, the capacity to ask: how do you know? And the instinct to distrust the fluent answer, and the patience to sit with the discomfort of not-yet-knowing, which is where all the real learning begins.
These capacities can’t be downloaded. They’re built slowly, through exactly the kinds of tasks that AI now makes it easiest to skip, and scientific studies are catching up with what educators already sense.
A landmark MIT Media Lab study tracked a group of students writing essays variously with ChatGPT, a search engine, or nothing at all. Those using LLMs showed the weakest brain connectivity of all three groups, cognitive activity scaled down in direct relation to how much was outsourced.
Most couldn’t recall what they’d just written and yet the task was completed. But fundamentally, the learning never happened. Worse still, the cognitive habits don’t automatically reset. Once we hand the thinking over, our brains don’t automatically take it back.
By the end of 2025, a RAND survey of 1200 students found two-thirds believed AI was harming their critical thinking. The students themselves can feel what’s happening to them. The most important thing educators teach has never been content.
It has always been periagoge, the reorientation of the whole person toward truth and the willingness to be wrong, to revise, to trace an idea back to its roots and ask whether it holds. If we design our curricula, our assessments and our institutions around the assumption that the output is what matters – the essay, the answer, the finished product – then we are not educating.
Plato’s escaped prisoner, having seen the sun, returns to warn the others. They don’t thank him. They find him disorienting, probably dangerous, certainly annoying. In their minds, the shadows they trust are sharper than any daylight he can describe.
Today, the cave is still the cave, but the chains are more comfortable than ever, and the shadows have learned to speak.
The question now is whether we are still teaching people to turn around and face the light.
Lucy Gill-Simmen does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
Original source: https://analysis1.mil-osi.com/2026/06/11/ai-doesnt-just-help-us-think-it-thinks-instead-of-us-what-this-means-for-the-process-of-learning/