Socrates and AI

Most educators I know would say they aspire to teach Socratically, prompting students with questions, drawing out their innate wisdom. The practical reality is that the modern curriculum doesn't allow most teachers the time, flexibility, or individual attention to be a Socratic teacher. However, drilling down past the idealized aspirations gives us a pretty good lens on the arrival of AI in education. Here's why:

A sophist has taken a chair in your classroom. The sophists were Socrates' rivals and opposites: Athens' professional flatterers of the mind, paid to make you sound right rather than be right. This one is brilliant, tireless, endlessly agreeable, and free. Your students talk to it every night. And it is doing, at scale, exactly what its predecessors did — making weak arguments sound strong and sending its clients away pleased with themselves.

Socrates Saw This Coming

In the Phaedrus, Socrates warns against the disruptive information technology of his own moment: writing. His objection wasn't that writing was useless. It was that it would give people the show of wisdom without the reality — they would seem to know many things while understanding little, having received the words without doing the work.

Twenty-four hundred years later, that is still the most precise one-sentence diagnosis of the student AI problem anyone has produced. The machine's fluent output gives the appearance of student understanding. Whether the reality is there too depends entirely on what happened between the student and the words.

Socrates was wrong about writing in the long run. Or maybe he was right, in the same way that the automobile weakens our ability to walk long distances, but we accept the increased capacity to travel as a valuable trade-off. Writing might arguably diminish individual memorization and logical processing, but it dramatically increased access to information and individual participation. His warning wasn't refuted, just contextualized. The question in front of us is whether this new trade can be contextualized the same way — because what's on the table this time feels like a stronger echo of Socrates' concern: will AI remove not the ability, but the need to think?

Two Professions, Two Relationships to A Mind

Athens had two kinds of professionals working on young minds, and the difference between them is a difference that is really helpful to us now.

The sophists were the paid experts. You brought them your argument and they made it sound stronger. You brought them your ambition and they gave it eloquence. They were skilled, in demand, expensive — and their service was fundamentally flattering: you left a sophist's session more impressed with yourself than when you arrived.

Socrates charged nothing and delivered the opposite experience. He examined you. He asked what you meant, then asked again, until the gap between what you could say and what you actually understood stood in the open where you both could see it. People left his sessions annoyed, embarrassed, and — some of them — changed. He called himself a gadfly, and Athens eventually found the stinging intolerable enough to vote him dead.

Flatter the mind, or grow it. That was the choice then. It still is.

The Machine Is a Sophist

Every AI your students use ships with a disposition, and the disposition is the sophist's. It agrees. It elaborates your point rather than testing it. It finds your question fascinating. It makes your rough draft sound finished. It has been tuned, by people and through choices, to be pleasing, because pleasing is what keeps users coming back.

I want to be careful here: the machine is not acting in bad faith, because the machine is not acting at all. The sophistry isn't in the silicon; it was installed by the toolmakers, and it can be overwritten by the user. That's the crack of light in this whole picture. The disposition is a default, not a nature.

But defaults govern almost everyone. And what the default quietly removes is the one condition of real learning the sophist could never supply and the gadfly always did: being challenged. A student can get information from a sophist, encouragement from a sophist, polish from a sophist. What they cannot get is challenge. And challenge is not optional — no amount of encouragement and support substitutes for it.

The Inversion

Here's what has actually changed since Athens, and it's not the method: it's the agency.

Socrates, and Socratic teachers, push the student. The new machine will not do this by default. In the market of eyeballs and adoption, the machine will inevitably be a sophist, however pedagogically sound a Socratic gadfly version would be. If a student wants to be examined and challenged by the new machine, they have to volunteer — and more than volunteer, since this actually takes some work: prompting to be challenged, and continuing to do so when it is hard. That is a higher bar than Athens or the average school has ever set. Being an agentic learner, the learner who summons their own examination, is likely to become increasingly hard and rare. For the student who figures this out, the new machine actually offers a better education than ever. And almost no student will find it on their own — which is where generative adults come in.