Writing as Thinking
In the Phaedrus, Socrates tells a story about the Egyptian god Theuth, who invented writing and presented it to the king as a gift for memory and wisdom. The king was unimpressed. Writing, he said, would produce forgetfulness in the souls of those who learned it, and the appearance of wisdom rather than wisdom itself. Socrates added his own objection: written words stand there with the solemnity of living beings, but if you question them, they maintain a majestic silence. They cannot answer. They cannot defend themselves. They say the same thing forever, to everyone, whether or not it is the right thing to say to this person on this afternoon.
Most people disagree with Socrates on this point (and often argue that if Plato had followed his advice and not written anything down we wouldn’t even be talking about it today. In a recent essay in The New York Times, Joe Cruz, a philosophy professor at Williams College, resurfaces this idea as he argues that universities have made a category error by treating the written essay as the highest cultivation of thought — and that the error is now so expensive, in the age of fluent machine prose, that it is worth revisiting. Natalie Wexler, who has spent years arguing that writing instruction is one of the most neglected levers in American schooling, replied with a defense of writing as fundamental to learning.
We think both of them are largely right, and that the disagreement is more useful than either position alone.
What the professor gets right
Start with the part of Cruz's essay that is uncomfortably true: the institutional response to AI in education has looked like an embarrassing arms race. Proctored blue books. Rooms shielded from the internet. Composition platforms that log every keystroke so that the word-by-word genealogy of a paragraph can be audited like a chain of custody. Cruz describes professors rejoicing over a student's typo as proof of human authorship.
His alternative is a throwback. He describes teaching without a script, running seminars as messy searching conversation. Last semester he let his students use AI on their final essays under stated guidelines, then sat each of them down for a forty-five-minute oral examination on their own claims. The result reassured him. The students were too quick on their feet, too deft at reconfiguring and extending their arguments, for him to believe the ideas weren't theirs. And — this is the line worth pausing on — at least half the class was more agile, more reflective and clearer in the oral exam than on the page.
Anyone who teaches one-to-one will recognize this immediately. We see it constantly at Cicero: There’s no place to hide in conversation. And, if the teacher is skilled at creating the right context and environment, the conversation can perform all the magic that Socrates claims writing destroys.
Cruz ends with a strong claim for what conversation can do for our overall thinking:
Consider what we might gain if we practiced speaking more and carrying around the resources of mind through our own attention, memory and engagement with one another. We may end up spending less time isolated, gain more practice resolving disagreement, and find ourselves a clearer sense of what we believe.
Where he is too casual
Cruz claims that the evidence for writing-advances-thinking is thin, and anchors this claim in a critique of the the writing-across-the-curriculum movement of the 1970s and British education theorist James Britton's 1980 hypothesis that we discover what we think in the act of writing. A hypothesis, Cruz notes, never proved.
Two problems. The first is historical. The idea that composing an argument in continuous prose is how you learn to hold an argument is not a product of the seventies. It dates back earlier to the progymnasmata of the rhetoricians like Marcus Aurelius (who did a lot of writing!); to the medieval disputation, which was often written out; and to the the Oxford and Cambridge supervision, which for a century and a half has run on a weekly essay read aloud and then torn apart in conversation (and on which we base our model of one-to-one learning). Britton gave a modern name to an extremely old practice. Dating the claim to 1980 makes it look like a fashion rather than something we've been doing for an extremely long time.
The second problem is the standard of evidence. "The research is inconclusive" is doing an enormous amount of work in a sentence that then recommends replacing essays with oral examination. Don’t get us wrong: we love oral examinations and learning through conversation. It's critical to what we do at Cicero. But, we love writing as much!
Wexler's argument, which she makes in a serious body of work comprising her Substack Mind the Gap, is that writing fails in American schools not because the premise is wrong but because the instruction is weak: students are asked to produce essays without the knowledge to have anything to say, and without ever having been taught to build a sentence that can carry a complex idea. The failure of badly taught writing is not evidence against writing.
And there is the small matter that Cruz made his case in roughly a thousand words of carefully composed prose in a national newspaper, because that was the way to make it land, be examined, and be answered — by Wexler, among others, in writing. Socrates' objection was that text cannot defend itself. The last month has been a demonstration of text defending itself rather well.
The best curriculum has both
The honest conclusion is that Cruz has diagnosed a real disease and proposed a cure that is half a cure. And the missing half is the thing he is arguing against.
A student writes an essay not because the essay is an archive and not so that a grade can be anchored in something tangible, but because writing forces them to give shape to their ideas in a very specific way. It is an assertion with a shape that is, ideally, carefully crafted and iterated upon. In a sense, the student is in dialogue with themselves and with a text as they craft a great piece of writing. (We explore some of this and the ideas of Zena Hitz in this piece on reading as a social act.)
Then, in this ideal curriculum, the student sits down with a teacher who has closely read their work and, line by line, they pull it apart, develop it better, and defend it. The conversation does the work that's right for conversation, while writing holds it's place.
This is not a new invention. It is the tutorial or supervision, and it is the core of how we teach at Cicero: one student, one teacher, a piece of writing on the screen between them. The weekly essay is not the assessment. It is the occasion. What changes a student over a term is not the essay and not the conversation but the loop — write, defend, see where it broke, write again knowing what breaking feels like.
An education that drops either half will produce a half of a learner. Drop the conversation and you get students who can perform an argument on paper and cannot hold one across a table. Drop the writing and you get students who are quick, charming, plausible, but lack the ability to quietly contemplate and shape an argument on their own.
We should not want either. The machines, as it happens, are extremely good at producing the second.
Further reading
- Natalie Wexler, "Yes, Writing Does Advance Thinking", Minding the Gap.
- Joe Cruz, "What if we are mistaken in treating writing as the highest cultivation of thought?", The New York Times — the opinion essay that prompted the exchange.
- Plato, Phaedrus, 274c–278b — Socrates on Theuth, memory, and the silence of written words.