What Were School Essays For, Anyway?

Reflections on the value of writing in the age of large language models

What exactly are we losing if the essay stops being our primary means of evaluating students – aside from quizzes, oral exams, and lab work, that is?

ChatGPT and other large language models have made the idea of the take-home essay obsolete. According to people I know who teach, and comments I see around the Internet, this last term of school has been hellish because of fraudulent AI detection systems put out by the plagiarism detection companies. Teaching assistants have been contributing to internet comedy by posting screenshots of these systems judging, with >90% confidence, that the Declaration of Independence, the first chapter of the Book of Genesis, and the play within a play in Hamlet are AI generated. All this chaos is going to be temporary, however. Educators will settle on a new paradigm for evaluating whether their students are learning – though I hope it is something better than quill pens on parchment under the watchful, judging eyes of teaching assistants making less than minimum wage.

Before we can figure out what we’re losing, if anything, with the demise of the undergraduate essay, we need to answer what essays were for in the first place.

Why Essays?

Let's start from the cynic's point of view. There's a quote from Peter Drucker to the effect that "what gets measured gets managed." That means if you can put a measurement on something, you can track it and improve it. I often think of the cynical inversion of this, which is "what is easy to measure, gets measured.”

The essay exists, the cynic would say, because it is easy to mark. I'm not mocking the efforts of teaching assistants everywhere, wading their way through hundreds and hundreds of pages per term of leaden, baffling, atrocious undergraduate writing, having to patiently make comments and suggest points for improvement. And that’s not even counting putting up with the whining and grade grubbing they then have to endure during their office hours after they hand the marks back.

(Maybe marking essays written by ChatGPT would be a nice change?)

But for all its drudgery, marking essays is known task. You sit down with your stack of student essays and your grading rubric, you read, you mark it up with your red pen: Was it well-written? Even more basically, was it coherent? Did it make a continuous argument? Did it show knowledge of the subject? And then you get into the nitty-gritty details like spelling and grammar, sentence structure, argument form. I’m sure there are teaching assistants experimenting with feeding their courses grading standards into ChatGPT and letting it wade through the undergraduate prose. ChatGPT marking papers written by ChatGPT seems like a nice closed loop.

“Go read a book about X, then write me an essay about X on topic C.” Such has been the structure of midterm and final assignments for generations. In later grades, they make it more challenging by asking you to pick your own topic, and then it's a three-dimensional optimization problem for the student:

X = can I write it?

Y = how hard will it be to do?

Z = will it get me a good grade?

So, in light of all this, we have to ask ourselves: What is the true value of the essay as a form of assessment? If AI-generated content can pass plagiarism detection systems, and if AI can even grade essays based on a rubric, then what is the role of the essay in education? Is it merely a convenient tool for evaluation, or does it serve a deeper purpose in fostering critical thinking, creativity, and effective communication?

There's no reason to believe that just because we've been doing something one way for a long time, it's the best way to do it. Inverting Drucker's principle: if something is easy to show progress on, it becomes the target and the measure of success. Witness the addiction of educational reforms to regular, multiple-choice tests.

Eliezer Yudkowsky, in his book Inadequate Equilibria, takes a look at the medical profession and asks: Why has this profession—general practitioners, particularly—not been unbundled? That is, why is the person who diagnoses your condition the same person who comes up with a treatment plan for your condition and also the person who administers the treatment?

Why can't these be separate specialties? Have a class of diagnosticians whose whole job is to diagnose patients and who are compensated for their ability to do so accurately and penalized if they do it incorrectly. Have a class of treatment planners who take the diagnosis and work out an optimal plan of treatment, applying all of their knowledge to finding out what the best treatment plan is for a condition. And then have people whose job is administering the treatment, people who know everything about the effects of particular drugs, surgeries, and radiation therapies, and specialize in adapting them to the individual patient’s circumstances.

But these roles are all bundled together in medicine. Of course, there are specialties, but even the specialist often is the one asked to do the diagnosis and then come up with the treatment plan and then administer it. But no one, Yudkowsky points out, is going to unbundle these roles because every single person who gets through medical school can say, "Well, that's the way I learned it, and that's the way to teach it." They know how to teach it because it is the way they were taught and it is how they practice. To imagine reconfiguring everything would require a change in incentives, and no one has the incentive to change it—except patients, who might prefer more accurate diagnosis, more tailored treatment plans, and more responsive treatment plans than “come and see me in a month for another rushed ten minute conversation and new prescription.” But patients have always been a silent group with respect to the structuring of medicine.

So there is no reason to think that essays, which have grown up as a means of assessment particularly in the humanities and social studies, are the best way to do things. Just as the medical profession has bundled roles that could potentially be separated for greater efficacy, the education system has relied on essays as a default assessment tool. But is this the most effective way to evaluate students' understanding and critical thinking skills? As we consider the future of education, it's worth exploring alternative methods of assessment that may better serve both educators and students.

Alternatives

As an alternative to the set essay, I think project and case-based evaluations would be better. This is how education is handled in business, law, and medical schools: you’re presented, often as a a group, with a problem to solve, with defined limits on what you can do, both given by the problem scope and by what you’ve learned so far. Rather than write an essay on Drucker’s ideas about measurement and management, for instance, you’re given the case history of a company that is falling behind in production and expected to articulate, in class in front of the professor and your peers, a potential solution using Drucker’s ideas – where it makes sense. Law students are given case histories and expected to discuss and debate potential arguments for the defense or prosecution. Medical students are given descriptions of patients and their symptoms, and expected to apply what they’ve learned to both recommend tests and develop a diagnostic hypothesis.

But those are hard to set up, and not amenable to a one-hundred student, Intro to English Literature class. It’s an evaluation technique for small classes, with more intimate and regular teacher-student interaction. Not the thing for mass-produced undergraduate education. That requires a set form, with some predefined standards, and a minimizing of individual judgment. Thus, the essay.

So if there is a better way (though hard) and the essay is too prone to cheating to be worth preserving, what reason is there to keep it?

Articulation

Here we need to look from undergraduate classrooms to corporate America. Jeff Bezos has a No PowerPoint rule at Amazon. That is, for managers all the way to vice presidents, you’re not allowed to get by of a gift for gab and slick graphic design. I imagine he doesn’t care for TED talks, either.

To propose an idea at Amazon, or deliver a report. you must write a full essay—an essay that answers a set of required questions. And these are not five paragraph high school English essays either; the requirements specify they have to be five to seven pages long.

The really clever part of this rule is this: at the meeting where other managers and VPs discuss the proposal or examine the report, the first half of the meeting is dedicated to reading: each person has a copy, and they sit there together and read in silence, making note and annotations, and then they discuss it. How I wish that my seminar classes in university, or meetings I have attended in my working life, had this rule. There is no room for pretending you were not able to read the essay: No one pretends to read it; no one says, "Oh, I'm sorry, I was so swamped, I didn't have time." The meeting makes time for you to do it.

The reason for doing this, Bezos has said, is that it exposes sloppy thinking. Sloppy thinking stands out on a page, whereas presentations can lead to talking around problems or hand-waving away difficulties. People with charisma and presentation skills can get away with a lot. But an essay is a separate thing from the person who wrote it, and has to stand or fall on its own merits.

Some people, former Amazon employees, have complained that this technique rewards people who are verbally articulate. But here's the thing: if you can’t articulate, you can’t think.

Tim Urban, in a post on Wait But Why about Elon Musk’s company Neuralink, talks about a possible application of the direct brain interfaces. As someone who has always felt inarticulate, Urban dreams that he could hook himself up to a Neuralink device and directly download what is in his head to the page.

If direct brain interfaces ever happen, it’s not going to be like that. Here’s a secret you learn from writing a lot. You actually don’t have great ideas in your head – it only feels like that. You find ideas as you create.

Dan Simmons called it the "I've got a great idea" fallacy. As a writer of fiction, he is regularly accosted at parties by people who tell him, "I've got the greatest idea for a novel. Would you listen to it?" And Simmons quietly hears them out. Then, depending on how many drinks he's had, he may break to them the fact that a short story—a simple short story, a very straightforward short story like "An Occurrence at Owl Creek Bridge" by Ambrose Bierce or "The Tell-Tale Heart" by Edgar Allan Poe—isn't one good idea.

A good story is thousands of good ideas.

You can summarize the conceit of “An Occurrence At Owl Creek Bridge” in a sentence, which makes it seem like the story is one good idea. But it is thousands upon thousands of decisions – of generate and test cycles, for my readers who have studied artificial intelligence. You have to find your way forward from a starting point and a vague concept of the goal. Often this is an emotional impression of what the story should make someone feel, a view of a goal you’d like to reach.

When you start to write, you are exploring. I can’t emphasize the point enough, so I’m going to make it a heading unto itself:

Creativity IS Search

All creating is exploration and mapping of new territory.

Creativity is search. It's about venturing into the unknown and discovering new paths, new connections, and new insights. It's about making sense of complexity and finding patterns in the chaos. Beethoven's Seventh Symphony, as I mentioned earlier, is a kind of magic spell—a series of choices and decisions that, when brought together, create something transcendent and timeless. It's a map of possibilities that Beethoven charted through his own creative search, and it's a testament to the power of human imagination.

I recall an anecdote, I believe from Douglas Hofstadter—I'm not going to bother looking it up—about an exam given in to Chinese doctoral students at the end of their degree. You’re put into a room with writing tols and a stack of paper, given a few hours, and a single command: "tell us what you know."

The only way to fail this exam is either A) not to do it, or B) just to regurgitate facts—you're expected to synthesize what you've learned and demonstrate it. And you have to really know your stuff, because you only have a few hours in which to do it.

In this sense, the essay is not just a means of assessment or a way to demonstrate knowledge; it's a tool for critical thinking and self-reflection. It's a way to organize your thoughts, to clarify your arguments, and to communicate effectively with others. And in a world where AI and technology are reshaping the way we learn and work, the ability to think critically and articulate our ideas clearly is more important than ever.

So, while AI-generated essays may challenge the traditional role of the essay in education, they also highlight the enduring value of clear and thoughtful writing. Whether we're writing essays, proposals, or presentations, the act of putting our ideas into words is a powerful exercise

This idea of creativity as search applies not just to music, but to all forms of artistic expression. Whether we're writing an essay, composing a symphony, or painting a masterpiece, we're engaged in a process of exploration and discovery. We're making a map of a path through possibility space.

Look at any artist's sketchbook, any composer's notebook, or any writer's drafts, and you'll see evidence of this process. You'll see the paths that were explored and abandoned, the dead ends and detours, the revisions and refinements. You'll see the struggle to find the right words, the right notes, the right images—to give form to the intangible and to express the inexpressible.

And here's an interesting corollary: everything we have in this world is suboptimal. We don't have the best "Paradise Lost," the best "Quake" software, the best Seventh Symphony, the best "Maltese Falcon," or the best "Citizen Kane." There is, somewhere out in the space of possibility, a better version that wasn't instantiated. But we should be grateful for what we have, for the creative works that have enriched our lives and shaped our culture.

So this is what I think is lost if we don't keep the essay around—if it drops out of education. I'm not saying it shouldn't be adapted or that it shouldn't be supplemented with other forms of assessment. But the essay has a unique value as a tool for critical thinking and self-expression—a way to engage with the complexities of the world and to articulate our own understanding.

The essay helps us articulate our thoughts to actually find out what we really think. It's why I think writing is so valuable, and even if you don't publish it, I would encourage everyone reading this to write, if only for yourself. Write longhand, write on a computer, dictate and then transcribe your notes using one of the readily available, essentially free speech-to-text programs, like those bundled with Microsoft Office. You'll gain a lot from it. And it's a shame that we don't teach it this way in school.

The essay is often presented as this ancient, archaic thing you have to do, which is going to be of no relevance to you as an accountant, as an engineer, or as a chemist. But it has this incredible personal value—figuring out what you think. I cannot tell you the number of ideas that have come to me because I actually sat down and wrote things out. Complete revisions, totally new ideas, radical new approaches—I actually have to keep a notepad nearby when I'm writing so I can jot down all the tangent ideas and other things to do.

So go out and write, and don't stop writing, even if the machines seem like they can do it better for you. Because while AI and language models may be able to generate text that mimics human writing, they cannot replicate the unique insights, experiences, and perspectives that each of us brings to the page. Writing is not just about producing words; it's about exploring our own minds, making sense of our experience. And that is something machines cannot do for you.

GPT-5, -6, … -N will certainly be able to, with the right prompt and random seed, write what you can, as well or better. But they cannot write what you - here, now, Being-In-Time - will create.