Two essays on writing first drafts
Taxonomy of generative AI usage
This table is based on this document from Illinois State University. “GenAI” is short for “generative AI,” which is an umbrella term that covers both large language models (LLMs) for generating text (e.g., ChatGPT) and image generators (e.g., DALL-E).
The table isn’t immediately relevant to the first reading, but for our discussion it’ll useful to have it at the top of the page.
| Level | Level descriptor | What this looks like in practice |
|---|---|---|
| 0 | No use of GenAI | Students will create their own, original work without the use of GenAI in any manner. |
| 1 | Organization and summarizing | Students will create their own, original work without the use of GenAI; however, the use of GenAI for personal efficiency (i.e., summarizing notes/readings, clarifying content) is acceptable. |
| 2 | Brainstorming | Students can consult GenAI as a tool for brainstorming or idea generation, but are expected to create their own, original work without the use of GenAI. |
| 3 | Feedback | Students create their own work, then use GenAI as a tool to provide feedback on their work. Students are expected to use feedback from GenAI to conduct their own revisions of their own work, so any work submitted should be GenAI-supported, not GenAI-created. |
| 4 | Co-creation and revision | Students can use GenAI to develop drafts/outlines of their work but are expected to carefully edit and revise GenAI-created content as appropriate for their learning context. It is expected that any use of GenAI-created content is properly disclosed and attributed. |
| 5 | Unrestricted, attributed use | Students can freely use GenAI if the use of any GenAI-created content is properly disclosed and attributed. |
| 6 | Unrestricted, unattributed use | Students can freely use GenAI in any form. Attribution is not necessary. |
Messy First Drafts
Now, practically even better news than that of short assignments is the idea of shitty first drafts. All good writers write them. This is how they end up with good second drafts and terrific third drafts. People tend to look at successful writers who are getting their books published and maybe even doing well financially and think that they sit down at their desks every morning feeling like a million dollars, feeling great about who they are and how much talent they have and what a great story they have to tell; that they take in a few deep breaths, push back their sleeves, roll their necks a few times to get all the cricks out, and dive in, typing fully formed passages as fast as a court reporter. But this is just the fantasy of the uninitiated. I know some very great writers, writers you love who write beautifully and have made a great deal of money, and not one of them sits down routinely feeling wildly enthusiastic and confident. Not one of them writes elegant first drafts. All right, one of them does, but we do not like her very much. We do not think that she has a rich inner life or that God likes her or can even stand her. (Although when I mentioned this to my priest friend Tom, he said you can safely assume you’ve created God in your own image when it turns out that God hates all the same people you do.)
Very few writers really know what they are doing until they’ve done it. Nor do they go about their business feeling dewy and thrilled. They do not type a few stiff warm-up sentences and then find themselves bounding along like huskies across the snow. One writer I know tells me that he sits down every morning and says to himself nicely, “It’s not like you don’t have a choice, because you do – you can either type, or kill yourself.” We all often feel like we are pulling teeth, even those writers whose prose ends up being the most natural and fluid. The right words and sentences just do not come pouring out like ticker tape most of the time. Now, Muriel Spark is said to have felt that she was taking dictation from God every morning – sitting there, one supposes, plugged into a Dictaphone, typing away, humming. But this is a very hostile and aggressive position. One might hope for bad things to rain down on a person like this.
For me and most of the other writers I know, writing is not rapturous. In fact, the only way I can get anything written at all is to write really, really shitty first drafts.
The first draft is the child’s draft, where you let it all pour out and then let it romp all over the place, knowing that no one is going to see it and that you can shape it later. You just let this childlike part of you channel whatever voices and visions come through and onto the page. If one of the characters wants to say, “Well, so what, Mr. Poopy Pants?,” you let her. No one is going to see it. If the kid wants to get into really sentimental, weepy, emotional territory, you let him. Just get it all down on paper because there may be something great in those six crazy pages that you would never have gotten to by more rational, grown-up means. There may be something in the very last line of the very last paragraph on page six that you just love, that is so beautiful or wild that you now know what you’re supposed to be writing about, more or less, or in what direction you might go – but there was no way to get to this without first getting through the first five and a half pages.
I used to write food reviews for California magazine before it folded. (My writing food reviews had nothing to do with the magazine folding, although every single review did cause a couple of canceled subscriptions. Some readers took umbrage at my comparing mounds of vegetable puree with various ex-presidents’ brains.) These reviews always took two days to write. First I’d go to a restaurant several times with a few opinionated, articulate friends in tow. I’d sit there writing down everything anyone said that was at all interesting or funny. Then on the following Monday I’d sit down at my desk with my notes and try to write the review. Even after I’d been doing this for years, panic would set in. I’d try to write a lead, but instead I’d write a couple of dreadful sentences, XX them out, try again, XX everything out, and then feel despair and worry settle on my chest like an x-ray apron. It’s over, I’d think calmly. I’m not going to be able to get the magic to work this time. I’m ruined. I’m through. I’m toast. Maybe, I’d think, I can get my old job back as a clerk-typist. But probably not. I’d get up and study my teeth in the mirror for a while. Then I’d stop, remember to breathe, make a few phone calls, hit the kitchen and chow down. Eventually I’d go back and sit down at my desk, and sigh for the next ten minutes. Finally I would pick up my one-inch picture frame, stare into it as if for the answer, and every time the answer would come: all I had to do was to write a really shitty first draft of, say, the opening paragraph. And no one was going to see it.
So I’d start writing without reining myself in. It was almost just typing, just making my fingers move. And the writing would be terrible. I’d write a lead paragraph that was a whole page, even though the entire review could only be three pages long, and then I’d start writing up descriptions of the food, one dish at a time, bird by bird, and the critics would be sitting on my shoulders, commenting like cartoon characters. They’d be pretending to snore, or rolling their eyes at my overwrought descriptions, no matter how hard I tried to tone those descriptions down, no matter how conscious I was of what a friend said to me gently in my early days of restaurant reviewing. “Annie,” she said, “it is just a piece of chicken. It is just a bit of cake.”
But because by then I had been writing for so long, I would eventually let myself trust the process – sort of, more or less. I’d write a first draft that was maybe twice as long as it should be, with a self-indulgent and boring beginning, stupefying descriptions of the meal, lots of quotes from my black-humored friends that made them sound more like the Manson girls than food lovers, and no ending to speak of. The whole thing would be so long and incoherent and hideous that for the rest of the day I’d obsess about getting creamed by a car before I could write a decent second draft. I’d worry that people would read what I’d written and believe that the accident had really been a suicide, that I had panicked because my talent was waning and my mind was shot.
The next day, I’d sit down, go through it all with a colored pen, take out everything I possibly could, find a new lead somewhere on the second page, figure out a kicky place to end it, and then write a second draft. It always turned out fine, sometimes even funny and weird and helpful. I’d go over it one more time and mail it in.
Then, a month later, when it was time for another review, the whole process would start again, complete with the fears that people would find my first draft before I could rewrite it.
Almost all good writing begins with terrible first efforts. You need to start somewhere. Start by getting something – anything – down on paper. A friend of mine says that the first draft is the down draft – you just get it down. The second draft is the up draft – you fix it up. You try to say what you have to say more accurately. And the third draft is the dental draft, where you check every tooth, to see if it’s loose or cramped or decayed, or even, God help us, healthy.
First Drafts in the AI Era
How will people compose text moving forward, now that every author working with a digital word processor and internet access can use generative AI? Many will likely opt to write traditionally as they did before, but some will use AI in partnership to draft. At this point, the methods a writer uses to develop a first draft feel like a dealer’s choice dilemma—ask AI to generate the draft for you, or bring some of your writing to the table and have AI help expand and refine it. If students use AI in their drafting process, I’m increasingly drawn toward advocating for the latter method.
Why First Drafts Are Crucial
I don’t like the idea of students going to AI and prompting a first draft. I know some have argued that this could be a helpful method to fight the blank-page anxiety most writers feel. Others view this as helping maturing writers by giving them a template or outline to help them organize and scaffold their ideas. I think there may be some value in those approaches, especially in terms of helping struggling students who might otherwise balk at writing, but all of these approaches assume a maturing writer will then use their budding rhetorical knowledge, content knowledge, and contextualize knowledge to complete the draft. Those of us who’ve taught first-year writing likely raised a questioning eyebrow at that idea.
Students struggle quite a bit when writing. For many, that struggle is a productive one, helping them exercise habits of thinking and self-inquiry, testing ideas, taking creative risks, and often failing. Anne Lamott’s “Shitty First Drafts” lays bare this process with frank elegance. I wish developers of LLMs would read it because as Lamott puts it, there’s a profound disconnect in how many fail to divorce the reality of the writing process from the end product:
People tend to look at successful writers who are getting their books published and maybe even doing well financially and think that they sit down at their desks every morning feeling like a million dollars, feeling great about who they are and how much talent they have, and what a great story they have to tell; that they take in a few deep breaths, push back their sleeves, roll their necks a few times to get all the cricks out and dive in, typing fully formed passages as fast as a court reporter. But this is just the fantasy of the uninitiated.
That fantasy of the uninitiated doesn’t see the often maddening process that goes into shaping and forming the words and sentences on the page. Lamott does a wonderful job of articulating this struggle and demystifying it:
The first draft is the child’s draft, where you let it all pour out and then let it romp all over the place, knowing that no one is going to see it and that you can shape it later. You just let this childlike part of you channel whatever voices and visions come through and onto the page.
Having AI generate a draft for you might not give you the same experience because an AI-generated draft might be something’s child’s draft, but not your own. The creative risks aren’t the same. Yes, you get a fast start from going to a blank page to a rough draft but having a machine do that for you feels like leaping from the crowd in front of the Boston Marathon and sprinting the last hundred yards ahead of everyone else.
We should advocate for keeping as much of the productive struggle of writing in the process. Others will see this differently and make compelling arguments of their own, and I hear you, but at some point we need to find common ground between the increasingly polarized camps entrenching themselves around AI.
As I read this section, Watkins is giving at least two different arguments against using a LLM to generate a first draft. One is “having a machine do that for you feels like leaping from the crowd in front of the Boston Marathon and sprinting the last hundred yards ahead of everyone else.” The other is inspired by Lamott’s essay.
Before you move on, try the following three things:
Write out Watkins’ “Boston Marathon” argument by analogy more explicitly. How is writing a first draft like running a marathon? How is having a LLM generate a first draft like skipping most of the marathon?
Watkins’ other argument — the one inspired by Lamott — is based on the value of writing the “child’s draft.” Write out this argument more explicitly.
Copy the text above this note into ChatGPT or some other LLM, and then have it unpack the arguments using the previous two steps as prompts. How does its output compare to your answers? How does the activity of using the LLM to analyze the arguments (including writing an initial prompt, reading the output, and any followup prompting) compare to the activity of analyzing the arguments in writing?
AI after the first draft
I’m not suggesting we toss AI out the window entirely. Far from it. Once that first draft is down—once a writer has wrestled with their ideas and gotten them onto the page—that's when AI might become a useful tool. Google’s NotebookLM could help students connect the dots, organize their thoughts, and even discover new angles they hadn’t considered.
Generative AI works best if you give it specific data to hone and shape its output. After a student has a first draft, AI can be used as a powerful synthesizing and feedback tool to help students clarify and expand their ideas. Using NotebookLM as a writing aide uses generative AI to help shape a user’s ideas because it requires them to upload their words to use it.
A student could complete all the scaffolding portions of a major research essay organically: a research question; thesis, multiple child drafts, brainstorming notes, and even an annotated bibliography and upload those documents into NotebookLM and prompt the system to give them connections and outlines for putting it all together. Using AI in this way preserves a great deal of the thinking that went into the writing process.
Another idea is using a tool like NotebookLM to help students organize portfolios. My first-year writing students write snap reflections and longer in-depth unit reflections throughout the semester. When it comes time to put them together at the end of the term in a reflective portfolio, a student has dozens of messy moments that record what they’ve learned throughout the course. Many students are often so exhausted at this point that they don’t effectively synthesize or summarize their ideas into a compelling narrative that identifies how they learned. If they uploaded these to NotebookLM, then generative AI could easily create a Wikipedia page-like entry summarizing their learning, including footnotes that link back to individual reflections. In this instance, we’d advocate for students to treat it like a source on Wikipedia—a great starting point that could illuminate their own self-discovery and serve as a scaffold for them to explore their personal reflective practices throughout the semester.
Incorporating AI into the writing process is a balancing act, for sure. We shouldn’t be spending our time making our students tech-savvy—we should teach them to mix traditional and new tools to explore ways to more deeply get in touch with their own creative processes and advocate for ways we believe best serve that journey. We want to train writers who can harness both the old and the new without losing that essential human spark that makes writing not just a skill, but an art.