I’ve been thinking a lot recently about how to help people get started with AI. Not the technical side — there are plenty of people better placed than me for that — but the more fundamental question of how to approach it. What mindset do you bring? What’s the skill that unlocks it? I’m seeing people feel daunted, or underwhelmed, or frustrated, but as with most bits of new technology: the barrier isn’t usually the technology. I’m finding this even more the case with AI.
As I’ve started experimenting myself I’ve actually gone back to skills I learnt in my undergraduate degree and I’m coming to think the skill we need isn’t computer science or data literacy or anything obviously modern. It’s philosophy.
Philosophy is never about the answers, that’s first and most disorienting thing philosophy teaches you is that the question is the work. Not a stepping stone to the answer — the thing itself. Socrates didn’t know more than everyone else in Athens. What he had was a method — precise, relentless, assumption-exposing questioning. He called himself a midwife: he didn’t create knowledge, he helped other people bring it forth through the right kind of dialogue.
Most people approach AI like a vending machine. You put in a request and expect an output. But the people who get genuinely remarkable things from AI are doing something closer to what Socrates was doing. They’re in dialogue. They’re asking precisely. They’re aware of their own assumptions. They treat the question as the craft.
A philosophy degree, for all its apparent abstraction, is fundamentally a training in that craft.
One of the things I remember most vividly about philosophy tutorials was the focus on what was meant by every particular word. What do you mean by know? What do you mean by fair? What do you mean by real?
The point wasn’t pedantry. It was that vague language produces vague thinking — and that most arguments fall apart not because the logic is wrong but because the terms were never properly defined.
This transfers directly to AI. Vague prompts produce vague output. When you ask for something “interesting” or “good” or “professional,” you are handing the AI a problem it cannot solve without guessing. The philosophical habit of defining your terms before you proceed turns out to be exactly the right instinct.
Aristotle distinguished between three types of knowledge. Episteme: theoretical knowledge, facts. Techne: practical skill, craft, knowing how to do something. And phronesis: practical wisdom, judgment, the kind of contextual understanding that can’t be reduced to rules.
These map with surprising neatness onto the different things you might ask an AI to do. Do you want information? A task executed? A considered view on something complex? The prompt looks completely different depending on which you actually need — and most people never stop to ask themselves which one it is. It’s the difference between “what is the capital of France?”, “how should I structure this argument?”, and “what do you think I’m missing?” Aristotle, essentially, gave us the taxonomy.
Plato’s dialogues aren’t just records of philosophical content. They’re a demonstration that form determines what can emerge. A badly structured conversation simply cannot arrive at a good conclusion.
The same is true of prompts. What context does the AI need? What constraints should I set? What am I actually asking for? Good prompting is a compositional skill — you’re building a thinking space, and its shape determines what can happen inside it. Philosophy teaches you to treat structure as a meaningful choice, not just packaging.
Read the answer as carefully as you asked the question
The Sophists are the cautionary tale of the ancient world. Brilliant, fluent, and for hire, they could make any argument sound convincing — because they had separated rhetoric from truth. The goal was to sound right, not to be right.
AI can do this too. It produces fluent, confident, well-structured output that is sometimes nonetheless wrong. The philosophical habit of critical reading — not cynical, but rigorous — is more valuable now than it has ever been. Is this actually a good argument, or does it just sound like one? Are the premises sound? What’s missing? The Sophists didn’t disappear. They just got faster.
A note on how this post was written
I wrote this post in collaboration with Claude — Anthropic’s AI. I came to it with the idea, the personal thread, and the philosophical framework. Claude helped me develop the structure, draft the sections, and shape the language. I’ve edited and made it my own, but it was genuinely a collaboration.
But there’s something else worth saying. Claude also jogged my memory. The elenchus. The maieutic method. Episteme, techne, phronesis. Words I haven’t used in years but really aren’t at the front of my brain. The concepts hadn’t gone anywhere. They were just waiting. What I’d lost wasn’t the understanding, it was the vocabulary sitting on top of it.
I’m not sure how I feel about that. We talk about it as a tool for doing things — writing, summarising, analysing. But it’s also, a tool for remembering. I think we need to ask how reliable a tool it is and how we then go and check those memories.
We outsource so much memory now — contacts to our phones, facts to the internet. AI perhaps holds something subtler: the frameworks we once learned and half-forgot. The concepts stayed with me, even when the words didn’t.
What I did — bringing a clear question, a personal framework, and a sense of what I wanted — is exactly what this post is about. The philosophy degree came first. Claude came later. The thing that made the collaboration work was knowing how to ask.
As a final note, Claude was very positive in tone about this collaboration, I’ve added in more of a tone of skepticism back in as I feel uncomfortable and I need to think through why. It’s a powerful way to work (memory!) and I have been teasing this framework for ages and not getting anywhere so its really satisfying to get this out there, but I need to reflect on what I’ve learned form this and how its left me feeling.
Leave a Reply