After years of studying how human beings write, the cognitive psychologist Ronald T. Kellogg concluded that drafting a basic persuasive essay was as mentally taxing as digging a ditch was physically exhausting.
Writing well requires a command of vocabulary, spelling and grammar. It forces writers to comprehend a subject, to sift through their working memory of that topic, to organize and plan. Writing an essay is so difficult that a person in the throes of it can do hardly anything else — as anyone who has shushed a spouse or child while trying to compose a decent sentence can attest.
Very few people would volunteer to dig a ditch. Likewise, it is hard — almost impossible, many English teachers will admit — to turn a reluctant writer into an enthusiastic one.
Yet the American education system expects nearly all students to learn how to write an essay. It’s little wonder, then, that as soon as a tool arrived promising to lessen the pain people grabbed it.
Like it or not, artificial intelligence is now firmly embedded in how students produce text. Somewhere between two-thirds and 90 percent of high school and college students are already using A.I. for schoolwork, according to recent surveys, often requesting help with writing assignments.
The history is the best part of the piece, and it works against the conclusion. Written examinations replaced oral ones because the typewriter and the clerical economy created jobs that ran on text, and because oral testing collapses once class sizes grow. Those are instrumental reasons. Goldstein reports that writing’s workplace value is already declining, then defends writing on cognitive grounds instead. That is a retreat, and the article concedes it. The curriculum followed the economy once. Nothing in the piece explains why it will not follow the economy again.
The second move to watch is the conversion of difficulty from a cost into a benefit. Kellogg’s ditch-digging comparison establishes that composition carries a heavy working-memory load. The article then treats the load as the source of the value. Robert Bjork’s work on desirable difficulties gives mild support: some difficulties produce learning and others produce fatigue and avoidance. Digging with a spoon takes longer than digging with a shovel and teaches nothing extra. Graham says no part of the writing process can be handed to a machine without cost. That may be true. It leaves open which parts pay for themselves, which is the question a teacher deciding a syllabus has to answer.
Transfer is the weakest link and goes untested. The claim is that drafting an essay on The Crucible builds metacognition that shows up when the student evaluates a mortgage or a medical diagnosis. Educational psychology has promised that kind of transfer for Latin, formal logic, chess instruction, and working-memory training, and the results have been tiny each time.
The empirical base is thin. One small EEG study, still fresh and not yet replicated, showing lower brain activity and weaker recall among people drafting with a chatbot. A survey range of two-thirds to 90 percent for student use, a spread wide enough to suggest the measurement is bad. Then the phrase about a growing body of research, which covers a literature that is mostly correlational and self-reported.
The article also does not ask whether the American school essay is a good vehicle for thought. My experience is that teachers wanted me to say things they favored and punished me if I did not tell them what they wanted to hear. The freshman assignment Goldstein feeds to the chatbots asks for a childhood reading experience that shaped the writer. The five-paragraph theme, the thesis-and-three-supports structure, the personal-growth narrative: these forms exist partly because they can be graded at volume by underpaid instructors. A machine produces them fluently because they are formulaic. That tells us something about the genre. The students in Binney’s class who found the chatbot’s poem generic were reading the machine correctly and also, without knowing it, reading the assignment.
Oral pedagogies are the obvious counterexample and go unmentioned. Socratic dialogue, moot court, chavruta study, the medieval disputation. These build the argument-testing and counterargument-anticipating habits the article attributes to composition, and they do it out loud. The piece’s own history notes that American colleges trained ministers and lawyers through oral examination until the 1870s. Nobody claims those men could not think. If writing is a technology for thinking, it is one technology among several, and the question becomes which one a given student needs.
There is an unresolved tension in the recommendations. Graham grants that AI helps students with disabilities, non-native English speakers, and the roughly one-third of American students reading below basic level, for whom drafting is painful and exposing a draft to another person is worse. Those are enormous populations. The article then recommends more unassisted writing across the curriculum for everyone. Somebody absorbs that cost, and the piece does not say who.
The strongest thread gets three sentences. The small number of MIT subjects who wrote with a chatbot could not remember what they had written and felt no ownership of it. The person who cannot recall his own argument has not made one. Ownership, memory, and the sense that the sentences are yours might be the thing at stake, and they are separable from vocabulary, grammar, and job-market value.
The only defender of full AI adoption is an executive at the company selling the product, who calls students consumers, which is an easy quote to hand a reporter and a bad one to give. Both students quoted land where the reporter lands.
I like the Sydney two-lane policy: some assessments secured and in person, others requiring AI use. It survives because it does not depend on detection, and detection is a losing game. The cost is administrative. Proctored writing means rooms, time, and a return to grading handwriting.
One story the article misses. Writing instruction is expensive because responding to student writing is expensive. That is why the formulaic essay persists and why class sizes cap how much writing gets assigned. Machine grading is arriving faster than machine drafting has, and it will be cheap. If schools take up the recommendation to assign more writing, they will need a way to read it. Watch that, and watch who ends up reading the essays students hand in.
