A Stranger's LinkedIn Bio Sent Me to Hegel, and I'm Still Lost

A new contact on LinkedIn had this in the education section of his profile:

Carleton University
M. A. (Incomplete), Philosophy
1991 – 1993
Activities and societies: I got to the one-yard line and forfeited. My unfinished thesis on Hegel's critique of Kantian morality could have been a great one. But no regrets...

I know Hegel and Kant the way I know the names of rivers in countries I have never visited. I have no formal training in philosophy. It is a hobby, and these essays are where the hobby gets expressed. I read the line and felt the pull of something I did not understand. So I acted on a whim. I opened a conversation with a language model and asked it what the man had spent two years on.

Nothing was riding on it. I was entertaining myself, which is a perfectly good reason to ask a question, and one that has only recently become cheap. You can sit down with something that has taken in a vast share of what people have written, and ask it about anything, at any hour, for no reason at all.

The Plain Version

It read several encyclopedia entries and Hegel's own paragraph on the subject, then wrote me an explanation in plain English.

I still do not understand it.

The explanation was accurate as far as I can check, and the words were short. The trouble is that plain words can stand on ground I do not have. Universal law. Contradiction. Ethical life. Each is ordinary English, each means something particular here, and the explanation leaned on all three in its first paragraph as if I had been handed the keys. I read them the way I would read a menu in a language I never studied.

I have been on the other side of this for thirty years. I know AI the way that man knew Hegel. When I explain what a weight is, or why a model can sound certain and be wrong, I use words that feel plain to me and arrive as a menu in another language. In this conversation the machine had billions of weights to draw on and I had almost no context for the subject. The strangers I explain AI to are in my seat.

One Notch at a Time

In 1982 the linguist Stephen Krashen proposed that people acquire a language from input that sits just past what they already understand. He wrote it as i+1: where you are, plus one. Too easy teaches nothing, too hard teaches nothing, and the useful material is one step up. It is the idea LinguaMama is built on, and I explained it in What Is Comprehensible Input?. Nothing in it is specific to language. I think it holds for any body of knowledge: the next idea has to be built entirely from ideas the reader already has.

So I asked for a rebuild. Each rung may use only what the rungs below it supplied, and the first one has to be something I already own. Here is Kant and Hegel that way, at a size I can hold.

Start with what you own. Most of us heard "what if everybody did that?" as children. It is the retort to cutting in line, dropping litter, or skipping the work and enjoying the result.
Kant turned the retort into a test. Before you act, ask whether you could want everyone in your position to act the same way. If you could not, the act is wrong. His standard example is the lying promise. Suppose everyone borrowed money while privately planning not to repay it. Promises would stop meaning anything, and the lie would have nothing left to lean on. The act destroys the thing it depends on. He also held that you should do the right thing because it is right, and not because it pays. That is the "duty for duty's sake" the encyclopedia kept repeating.
A generation later, Hegel asked where the test gets its facts. The lying promise fails because promising already exists and you already want it to. The test protects things you value. It cannot tell you what to value. Take property. Hegel wrote that a world with no property is no more contradictory than a world without some particular nation or family. Stealing is a contradiction only if you have already decided that property should exist, and "a contradiction must be a contradiction of something." The test checks an act against something you already hold. It does not supply the something.
So where does the something come from? From the people who raised you: the family, the town, the laws, the way promises get kept and work gets done around you. Hegel called it ethical life. You did not derive fairness from a formula. You absorbed it from people who practiced it and corrected you, the way you absorbed your first language. That comparison is mine, not Hegel's, and I make it on purpose, because it is Krashen again: acquired, not studied.
The argument is not settled. Kant's defenders say Hegel was unfair. Kant offered other versions of the test, including treating people as ends in themselves and never merely as tools, and he never meant it as a calculator. Scholars are still arguing about it two centuries on, and I am in no position to referee.

That is five rungs, and I can hold them. I could not hold the first version. The ideas are the same ones. What changed was the order and where it starts.

The Other Ladder

The machine on the other end of the conversation needs the same treatment. For fun, I would call it a conversation with the sum of world knowledge, and that is not quite right.

You have typed a question into a chatbot and gotten a sentence back. That is where you are. Behind the sentence is a program that was shown an enormous amount of human writing — books, articles, forum posts, code. It was given one exercise to repeat billions of times: guess the next word.

What that exercise leaves behind is a very large set of numbers, the weights. It is not a library, and nothing gets looked up. You do not remember the conversations that taught you English; you simply speak it. The model keeps the shape of what it read, not the pages.

That is how it can answer a question about Hegel at any hour with the manner of someone who has read the field, and how it can be wrong in exactly the same manner. It is why I had it read the actual sources before it answered. It also knows nothing about you beyond what you type, as I argued in I Know Kung Fu. It did not know I was lost until I said so. And it will explain a third time as patiently as the first, because it has nothing at stake in whether I get it, which is a reason the checking stays with me.

The Stranger Is Me

The man with the unfinished thesis got to the one-yard line and stopped. I am nowhere near the one-yard line. I still do not understand Hegel.

What I understand better is what it costs to explain anything to a person who did not spend your years. Find the sentence they already own, start there, and add one idea per step, using only what the previous steps gave them. A plain word is not enough. It has to stand on something the reader already has. That holds for Kant, it holds for weights, and I suspect it holds for most of what I have said to strangers about AI.

A five-page summary of this argument is available as a PDF carousel.

This piece was drafted with Claude, which is made by Anthropic. Claude did the reading described in it.