Claude and his recommendations on classical education
For parents of children in primary and secondary school, in the age of AI.
1. The short version
The best idea in classical education is that a child should learn how to reason, not just what to believe. It matters more now than it ever has, because AI will hand your child fluent, confident answers to almost anything, and some of those answers will be wrong.
But the idea is easily oversold. Thinking skills are not a separate muscle you can train in the abstract. They grow out of knowing things. A child who has read widely, can do the maths, and has practised arguing and writing is far better placed than one who has only memorised the names of logical fallacies.
My recommendation: keep the goals of the classical tradition (clear thinking, clear writing, wide reading, sound arithmetic), keep a solid base of real knowledge, and add one new habit for the AI era: do the thinking first, then check it against the machine.
2. What to do, by stage
Primary school (about ages 5 to 12)
- Build the knowledge base. Reading, writing, times tables and a broad sense of history and the natural world. This is what later reasoning runs on.
- Read aloud and read together, well past the age they can read alone. Stories are how children first learn what other people feel and want.
- Ask "how do you know?" lightly and often. At dinner, in the car. Keep it curious, not like an interrogation.
- Let them struggle briefly before helping. Productive struggle is where the learning happens.
- AI at this age: keep it supervised. If you use it, use it with them, and treat it as something to question ("Let's see if that's actually true"), not as an oracle.
Early secondary (about ages 12 to 15)
- Teach simple argument skills. What is a claim? What is the evidence? What would change your mind? Common mistakes in reasoning (such as attacking the person instead of the point) are worth naming, but treat them as a way to check your own thinking first, and other people's second.
- Encourage writing and debate. Writing is thinking made visible. Arguing the side they disagree with is one of the best exercises there is.
- Set a rule for AI homework: attempt it first, then use AI to check, explain or challenge. Never to produce the finished work.
- Teach that AI can be wrong. Have them catch it in a mistake on purpose. It is a memorable lesson.
Senior secondary (about ages 15 to 18)
- Philosophy, logic and history are worth seeking out as electives, clubs or reading, if the school doesn't offer them.
- Practise checking sources. Where did a claim come from? Who benefits from it? Can you find it somewhere independent of the AI that told you?
- Make them explain their reasoning aloud without notes or tools. If they can't explain it, they don't yet understand it.
- Talk about motivated reasoning. All of us find flaws quickly in arguments we dislike and wave through those we like. Knowing this about yourself is more useful than any list of fallacies.
3. What AI changes
- Answers are now cheap. Judgement is not. The valuable skill is deciding whether an answer is right, and that needs knowledge of your own to compare it against.
- Doing the thinking first matters more. Early, small studies suggest that leaning on AI too soon can reduce how much people learn and remember. The research is young and I would not treat any single result as settled, but the direction is plausible and fits what we know about learning generally.
- Fluent is not the same as true. AI writes smoothly even when it is wrong. Children (and adults) tend to trust smooth writing.
- Used well, AI is a patient tutor. It can explain a concept five different ways, quiz your child, and argue the other side. That is real value, as long as your child is doing the work and not the machine.
- Honest effort still has to happen somewhere. If a child never writes a bad first draft, they never learn to write.
4. What holds up in the classical idea
- Logic, rhetoric (clear expression), history, philosophy, literature and mathematics all train the mind, and they are neglected in many places.
- Learning to recognise a weak argument is protective. It is harder to be misled by advertising, politics, social media or a persuasive AI.
- Literature and history teach what people are like, which no textbook of rules can replace.
- Education aimed at a good life, not only a job, is a worthy aim.
5. What doesn't hold up
- "Think before being told what to think" is not what actually happened. Classical education relied heavily on memorising texts and accepting authority, particularly in the early years. Independent questioning came later, if at all.
- It wasn't purely for life rather than work. Medieval universities trained clergy, lawyers and doctors. It was also an education for a small, mostly male elite.
- Thinking skills don't transfer as easily as people hope. As far as I know, the research (cognitive scientists such as Daniel Willingham have written about it accessibly) finds that people reason well mainly in areas where they have knowledge. General "critical thinking" taught on its own tends to carry over weakly.
- Knowing the fallacies can backfire. It can turn into a way of winning arguments, or of dismissing anyone you disagree with, rather than a way of finding out what's true.
- Modern schooling isn't simply "teaching what to think." Critical thinking is a stated goal in most curricula, though how well it's achieved varies a lot between schools and teachers.
6. What's worth taking from the article you posted
- The framing is good: a young person should learn how to think, not only be told what to think. It is a worthy goal even if the history is romanticised.
- Teaching epistemology (how we know what we know) every year, from kindergarten to university, is an interesting proposal. I know of no solid evidence it works, so treat it as a promising bet, not a proven fix. At home, the version of this you can do for free is the "how do you know?" habit.
- One parent's approach is worth copying: showing children directly that they are being taught to think, and demonstrating it, instead of just complaining about schools.
- Caution about the tone: lines like "trained seals" make for good posts but are not evidence. Be wary of any claim that flatters your existing view of schools.
7. What we don't know
- Whether classical-style schooling produces better thinkers, or whether families who choose it simply differ from the rest. I'm not aware of strong evidence either way.
- How AI use in childhood affects long-term learning. The studies are early and small.
- Whether a dedicated course in logic or epistemology beats simply reading widely, writing often and arguing well.
8. Five things to start this week
- Ask "how do you know?" about something at dinner.
- Agree a house rule: attempt first, then check with AI.
- Have your child find a mistake in an AI answer.
- Read something together that is old, hard and good.
- Have them explain something they've learned, aloud, without notes.
These are my views, based on what I know of the evidence and the history, and not a substitute for talking to your child's teachers.