Every word in this sentence is the result of choices.
One of the courses I teach with some regularity is our department’s course on Style. I base the first half of the course on the ideas behind Raymond Queneau’s Exercises in Style, a book where he takes a simple 2-page narrative and rewrites it in something like 125 different “styles.” The best thing about this approach to writing is that it really focuses a writer’s attention on the specifics of language. At a certain point, the narrative itself is irrelevant—the entire book is the same anecdote told over and over. Instead, language itself takes center stage, and that’s what I want for my students in the Style course. I expect them to pay closer attention to language than they likely would in any other course. I ask them to think about things like word order, parts of speech, the connotations that emerge from word choice, etc.; in other words, the course focuses on the features that are often invisible in our writing (unless those features go wrong and get flagged in feedback).
There is method to my approach. It’s an old book now, which had the misfortune of coming out just after 9/11, when talking about anything other than 9/11 just wasn’t tasteful. The book is Don Foster’s Author Unknown (it’s old enough not to have a spot in Bookshop.org). Foster described himself as a “literary detective,” and for a short while, he held a reputation far beyond that of most English professors. He was the person who figured out that the author of Primary Colors was Joe Klein and he worked on the Unabomber case and the JonBenet Ramsay kidnapping, among others. In many ways, his methods have been subsumed by digital text analysis, but at the time, it was pretty cool stuff. His book made the case that no two people write in quite the same way, that individual style was not unlike DNA. When we consider an author’s style, we tend to look at prominent surface features, the same way that we distinguish among people according to physical characteristics that are pretty general (size and shape, hair color, skin color, et al.). In the same way that we can’t see someone’s chromosomes, most of a writer’s intrinsic style happens beneath our notice. Writing style happens across a range of scales1.
One of my goals for Style, then, is to get students to spend more time at those smaller scales, to think about the micro scale choices they make when they write. One early exercise involves asking them to choose a single sentence and to rewrite it 10 different ways. Every word in this sentence is the result of choices can be versioned in dozens of ways: I consciously chose each word of this sentence. Intention lurks behind this sentence’s every word. The consequences of my selections appear in the words that compose this line. This sentence collects a bundle of decisions and presents them as though they were invisible. I could go on and on. Some of those choices are visible - do I identify myself as the person making choices? do I want to emphasize the words themselves or my decisions? can I get rid of that flat “to be” verb in the middle of the original? But some of them are tiny and largely inexplicable to me: why did I say “every word in this sentence” when I could just as easily have written about “each word of this sentence”?
My point here is that even a single, short sentence is packed with choices, from each individual word to the way that phrases hold together, to their collective ability to communicate a particular message. (And that’s to say nothing of the paragraphs, pages, and chapters in which a given sentence might be embedded.) But the sentences we end up writing, and others then read, are the tips of icebergs; the larger potentialities of those sentences remain below the surface, not just for readers but for ourselves as writers, unless we force ourselves to examine them more closely. The sentences and paragraphs that result are temporary stabilizations of the “glorious chaos” that Randall Munro describes above.
The “generative AI” debate perked up again last week with the publication of Ted Chiang’s essay (“Why A.I. Isn’t Going to Make Art”) in the New Yorker. You don’t need to get further than Chiang’s subtitle to see why I’ve emphasized choice as much as I have here; it reads that “To create a novel or a painting, an artist makes choices that are fundamentally alien to artificial intelligence.” Chiang’s discussion is framed in terms of art, and he explains how the interplay of scale is crucial:
What I’m saying is that art requires making choices at every scale; the countless small-scale choices made during implementation are just as important to the final product as the few large-scale choices made during the conception. It is a mistake to equate “large-scale” with “important” when it comes to the choices made when creating art; the interrelationship between the large scale and the small scale is where the artistry lies.
When those students of mine take their sentence and rewrite it in 10 different ways, one of the things they (often) discover is that the versions that result can’t simply be dropped back into their original text. And that’s part of the point, too. It’s not only the choices that we make when we’re in the middle of a sentence. Those decisions2 depend on what’s come before and they affect what follows. Any individual sentence might call back or foreshadow any other. When I rewrite a sentence here, as I do fairly regularly, the effect of that revision almost inevitably requires me to touch up the rest of the paragraph, and sometimes more.
One of the other implications of using Queneau’s book to undergird my Style course is the fact that his book explicitly uses the word “exercises,” which prompted me to consider the parallels between writing practice and physical exercise. Chiang explores this analogy as well:
The point of writing essays is to strengthen students’ critical-thinking skills; in the same way that lifting weights is useful no matter what sport an athlete plays, writing essays develops skills necessary for whatever job a college student will eventually get. Using ChatGPT to complete assignments is like bringing a forklift into the weight room; you will never improve your cognitive fitness that way.
There’s a lot to recommend the exercise analogy: it speaks to gradual improvement, the fact that change isn’t always obvious or immediate, the way that the ability to write well feeds into all sorts of other activities. When I talk to my students (or on this site) about “writing practice,” I mean something very much like exercise.
The problem is that reading and writing are tied, too often, to the end products rather than the process. This isn’t a problem in my own field, where “teaching writing as a process rather than a product” has been a mantra for decades, but I don’t know that we’ve always done the best job of explaining what that means to the other 99.9% of the world. And the infrastructure of education (class meetings, grade-able products, deadlines, assessment, course grades, et al.) mitigates against it. What value students gain from my courses doesn’t come from the products they generate; as Chiang explains, “teachers don’t ask students to write essays because the world needs more student essays.” We ask students to write essays, in part, because we all benefit from exercising those skills involved in translating our world into language. Those skills help us understand others’ translations (i.e., reading) and they overlap (albeit imperfectly) with the other ways we channel our selves and our worlds for and to each other.
From the outside, though, it’s possible to see an essay in isolation from the time, effort, and skill necessary to generate it. It’s possible to run it through an algorithm (as opposed to reading it), to reduce it to a bullet list of claims. And as we’ve learned, there are all sorts of tools that will offer to simulate them instantaneously. But while the exercise metaphor has its limits, the fundamental analogy at the heart of generative AI writing, that writing is nothing more than a sophisticated form of algorithmic autofill, is far more restrictive and corrosive. And that’s to say nothing of its other costs (economic, environmental, et al.) that we continue to ignore.
Postscript: I was all ready to hit the publish button on this post, but I was browsing through the day’s newsletters on Tuesday, and as so often happens, I was struck by something from Robin Sloan. This is part of a longer story about how he ended up in California, but he was sending daily emails to the person that would eventually hire him. And one of those contained the following:
Sloan is talking about the relationship between journalism and programming/coding (and 20 years ago), but this is a “solution” that is almost so pervasive in our culture so as to be invisible, sourcing work out to third-parties so that you don’t have to manage it yourself. I was thunderstruck by that third paragraph: If you send all your programming work to outside developers, what you’ll get, at best, is exactly what you wanted. And that’s a perfect analogy for what I want to say here about writing and AI: these corporations are begging us to outsource the act of communication, and there are even limited use cases where this makes some sense. But when we do so, when we fire up Gemini or Grammarly or ChatGPT, what we’ll get, at best, is what we ask for (and even that’s a gamble in a lot of ways). At best, we will get only what we ask for.
If good writing lets us see the tips of icebergs, and sometimes offers us glimpses of the “glorious chaos” that lies beneath the surface, AI-generated writing serves up ice cubes floating on an increasingly polluted sea. Writing takes practice and effort, yes, but the payoff is that chaos can result in the “cool solutions and inventions.” Art is one way to describe it, as are knowledge and wisdom. Those things don’t scale, but then, that’s a point in their favor, as far as we should be concerned.
And that’s where this post ends. I’ve got a bit more to say, from a different direction. More soon.
One great example of this is that each of us has a different “working vocabulary.” For example, one of the things that tipped Foster to Klein’s authorship of Primary Colors was the vocabulary overlap: “Some of the rarely used words that appear in 'Primary Colors' and Klein's writing include 'giddy,' 'gelatinous,' 'scruffy,' 'squishy,' 'lugubrious,' 'jittery' and 'tribal.'“
I’m using the word “decisions” here so as not to repeat the word “choices” from the prior sentence, for instance. It’ll let me use “choices” again in the next sentence if I want to, without sounding overly repetitive.