Using AI to Teach Us, Not to Think for Us

A philosopher's case for the tool, and the line it should not cross

There is a temptation, among people who care about careful thinking, to treat artificial intelligence as either a saviour or a threat, and to reach for one of these two postures rather than doing the harder work of figuring out where the tool genuinely helps and where it genuinely harms. I want to resist both postures here. AI is neither the end of independent thought nor its salvation. It is an instrument, and like any instrument, from the printed book to the calculator, its effect on the mind depends entirely on how it is used.

What the printing press can teach us about the current moment

It is worth remembering that every major leap in how information moves has provoked the same anxiety. When the printing press made books cheap and widely available, serious thinkers worried that memory would atrophy, that people would stop carrying knowledge in their own minds because they no longer had to. Socrates, centuries earlier, raised a related worry about writing itself in Plato's dialogue Phaedrus, suggesting that the written word would produce the appearance of wisdom without its substance because people would rely on external marks rather than genuine understanding held within.

He was not entirely wrong, and he was not entirely right either. Writing did change how human memory works, but it also made possible forms of thought, careful argument built across pages, ideas tested against a permanent record, that oral culture alone could never have sustained. The lesson is not that new tools are safe. It is that the question worth asking is never "does this tool change how we think," because it always does, but rather "does it change our thinking in a direction that serves understanding, or one that merely serves convenience."

Helping you to think or thinking for you?

Where AI genuinely teaches

Used well, an AI system can function as an extraordinarily patient tutor. It can explain a difficult passage of Kant three different ways until one of them lands. It can be asked to argue the opposing side of a position you hold, forcing you to test your own reasoning against a genuine counterargument rather than the weaker version of it you'd constructed in your own head. It can surface a source you didn't know existed, a philosopher you'd never encountered, a historical parallel that reframes a contemporary question. Used this way, it does what the best teachers have always done: it puts pressure on your thinking without doing the thinking for you.

Where it quietly begins to think for us

The danger creeps in at a specific, identifiable point: the moment you stop checking the tool's output against your own judgment and simply accept it because it arrived quickly and sounded certain. This is not a hypothetical risk. It is already a well-documented pattern, sometimes called cognitive offloading, in which reliance on an external system for a mental task gradually reduces one's own capacity for that task. The muscle that atrophies first is not knowledge, which the tool can still supply on demand, but judgment, the harder won ability to weigh, question, and decide for yourself.

There is a simple test I'd offer for the line between the two uses. After consulting an AI system on something that matters, ask yourself: could I now explain this to someone else in my own words, defend it against a good objection, and say honestly where I still have doubts? If the answer is yes, the tool has taught you something. If the answer is no, if you have only a polished paragraph you couldn't reproduce or defend, then you have outsourced the thinking rather than sharpened it, and the paragraph, however fluent, is not yet yours.

A discipline, not a prohibition

I do not think the answer is to avoid these tools, any more than the answer to the printing press was to avoid books. The answer is a discipline of use: treat the output as a first draft of thought rather than a conclusion, ask it to challenge you rather than simply agree, and always return, at the end, to the question of whether you yourself now understand something you didn't before, or whether you have merely borrowed the appearance of understanding for as long as the words stay on the screen.

The shamanic tradition I also draw from has a useful instinct here, though it predates any of this technology by millennia: wisdom that has not been earned through your own encounter with a thing tends not to stay with you. It passes through rather than settling in. The same, I think, is true of an answer an AI system hands you without your having wrestled with it first. Let the tool widen the door. But walk through it yourself.

Kathy Postelle Rixon is Chair of The Philosophical Society: Oxford, a researcher of philosophy of mind and physics, and shamanic practitioner. You can contact her at kathy@magicinharmony.com or visit her website at www.magicinharmony.com.