Presentations

Using AI to Teach Better

A short talk for teachers on bringing AI into a writing course without losing the process.

Added

Process: Directed drafting1 round of revision

When
For teachers · a faculty session
Length
About 20 minutes
Slides
15, each with its notes below it

Or press P. Arrow keys or a clicker move between slides, F goes full screen, and Esc comes back here.

For teachers · about 20 minutes

Using AI to Teach Better

Keep the process. Show your work, and ask for theirs.

Process: Directed drafting. The note at the end says who did what.

This is a talk for other teachers, about how I’ve brought AI into my writing courses. It starts where my own deck started: with a label for how the deck was made. Everything in it comes from what I already do in class.

AI User Gap in Education

  • Most students are already using AI before they get to my class.
  • Pretending they aren’t is academic theater.
  • The gap between people who use it well and people who don’t keeps widening.
  • Students don’t need a ban. They need practice using it ethically and well.

We’ve been here before. In the late ’90s and early 2000s, some professors insisted digital writing wasn’t “real” writing, while students were already composing on computers at home. Today everyone has a laptop in class. AI is the next literacy shift, and students who never practice with it in school leave behind the ones who did.

It doesn’t matter that much who starts the writing. What matters most is the recursive process: the work in between.

Murray (1972)
Teach writing as a process, not a product.

Sommers (1980)
Experienced writers revise the whole piece; students change words.

So
Look less at the first sentence and the last one, and more at the middle.

Composition moved from grading the product to grading the process half a century ago, after Donald Murray (1972). A lot of the worry about AI goes back to the product: who typed the first sentence, how clean the last one is. Nancy Sommers (1980) found that experienced writers revise globally and can’t always say where one draft ends and the next begins. That middle is where the learning is, with or without AI in the room.

One shot, or the loop

AI fits the recursive process. The trouble starts when a prompt or two replaces it.

Prompt Write a paragraph about a place that taught me something.

AI My grandmother’s kitchen taught me the value of patience, hard work, and family. Every Sunday, the smell of her cooking filled the house, reminding me that the best things in life take time.

1 pass · nothing decided, nothing to revise

  1. Student, a rough start my grandma’s kitchen. radio always on. she never measured anything and honestly it bugged me
  2. AI, a reader’s question The strongest part is “it bugged me.” Why did it bug you? Did that change?
  3. Student, a global revision I used to think my grandmother cooked carelessly. She never measured anything. It took me years to see she was tasting the whole time.
  4. AI, an editor’s suggestion Try cutting “the whole time.” Or keep it, if that’s the rhythm you want.
  5. Student, the call Keeping it. That’s the point of the paragraph.

Sample drafts, written by Claude

Press “One shot,” then “The loop,” and step through its rounds. The one-shot paragraph is finished and says nothing in particular, because nobody decided anything. In the loop the student and the tool take turns, the paragraph changes what it’s about in the third round, and the student makes the last call. You can work with AI on a text for a long time and make it good. Replacing that work with one prompt isn’t enough work, and it’s easier to spot.

Four ways of working

Every essay on my site carries one of these labels. Name the one in each scenario.

Scenario 1 of 4

You write a reading quiz yourself, then ask an AI tool to check it for typos. None of its wording ends up in the quiz.

Solo. A spelling and grammar check isn’t generative work, and none of it ends up on the page.

Scenario 2 of 4

You draft a rubric, paste it in, and go three rounds on the tool’s feedback until your criteria are clear. The sentences are still yours.

Draft-first. You wrote the draft; the tool was a reader and an editor over several rounds.

Scenario 3 of 4

You write the first half of an assignment sheet, the tool drafts the second half, and each of you revises the other’s part.

Co-writing. You traded passages, so the sentences are a blend, and so are the decisions.

Scenario 4 of 4

You talk through a lesson plan out loud. The tool drafts it, you tell it what’s wrong, and it revises until it says what you mean.

Directed drafting. The tool wrote most of the sentences; the ideas, the judgments, and the final approval are yours.

Sample scenarios, written by Claude

The four labels come from my essay on how I use AI. Solo: none of the words come from AI, though a spelling check or brief feedback doesn’t count. Draft-first: I write the draft, and the tool reads and gives feedback over several rounds. Co-writing: we trade passages and revise each other’s. Directed drafting: the tool writes most of the sentences from my ideas, and I direct every revision. None of them is the right way. Each is a different tool for a different piece, and in every one of them I’m responsible for what I publish.

The Open Kitchen Policy

  • I tell students when I use AI to make class materials, and how.
  • Students cite and disclose their AI use. That’s the line.
  • Nobody’s policing anybody. We show each other the work.
  • If I’m asking for their process, I show mine.

An open kitchen is one the diners can see into. I can’t teach what I haven’t practiced, so I use AI in my own writing and say how, on every page of my site. The same goes for class prep: a rubric, a prompt, an assignment sheet. Students see the label and know what they’re looking at. Asking students for transparency lands better when it goes both ways.

Write your kitchen note

Pick one thing you made for class this week. Say how it was made.

Process: Draft-first. I made this week’s discussion prompt with the tool and its model, on the date I made it. It gave feedback on two drafts. I wrote it, revised it, and made every call. I’m responsible for what’s here.

Stays in this tab. Nothing is sent.

Fill in the fields and the note writes itself: the way of working, the tool and its version, the date, what the tool did, and what you did. Copy it onto the assignment sheet or the course page. What you type stays in your browser tab and isn’t sent anywhere.

A citation says a tool was there, not what it did

MLA (2025)
Still cite it, and also describe how you used it.

APA (2025)
Cite the tool, and describe the use in the paper or an author note.

Weaver (2024)
The AID Framework: what the tool did at each stage of the work.

Full references in the notes below.

A citation treats AI like a source you consulted once, first or last. It can’t show the middle. The style guides have caught up: MLA still asks for a citation and adds a description of how you used AI (Modern Language Association, 2025), and APA asks you to describe the use in the paper or an author note (McAdoo et al., 2025). Kari D. Weaver’s (2024) AID Framework is a short statement that names the tools and says what they did at each stage. Lisa Ede and Andrea Lunsford’s (1990) two kinds of collaboration, one with fixed roles and one where the roles shift, are the vocabulary behind the difference between directed drafting and co-writing.

Grade the process

  • Full credit for a first rough draft, whatever state it’s in.
  • Workshop the drafts in class.
  • Read the next version for change, especially global change.
  • Students revise as many times as they need until the end of the term.
  • For AI, all I ask is that they cite and disclose it.

This is how I grade, and it’s why AI doesn’t worry me much. If the first draft gets full credit no matter what, there’s no reason to fake it, and the grade comes from what changes after it. We do enough rewriting by hand in workshop that I see the process happen in the room.

Hierarchies of Importance

Save your deepest focus for the work that matters most. Let AI take the rest.

← Deep human focus · Let AI handle →

  • Designing the syllabus
  • Grading final exams
  • Feedback on a student’s draft
  • Reformatting an assignment sheet
  • A reminder email about the due date
  • A citation guide for MLA and APA
  • A recommendation letter
  • Alt text for slide images

Deep human focus

    Let AI handle

      Not every text is worth the same amount of work. That’s a rhetorical decision every writer makes: I one-shot a caption, and I don’t one-shot a syllabus. Overthinking the small tasks eats the time AI saves. For me, designing the syllabus and grading final exams get the deepest focus. Sort your own week, and add tasks that aren’t on the list. Some of yours will land differently from mine, which is the point.

      Send me your sort

      Save the image and email it to helloeshoward@gmail.com. I’ll add your two lists to the site, where they stay unless you ask me to take them down. Your name and email address don’t go up.

      Tasks you added go up as you typed them, so leave out anything private. If one names a person or a school, I’ll make it general.

      Want Desk Copy, my monthly letter? Say so in your email, and I’ll add you within a couple of days.

      See what the room sorted →

      If you’d like your sort on the site with everyone else’s, save it and email it to me. On a phone, Save image opens the share sheet, so you can send it straight from there. Sending it is your choice, and it’s the one thing in this talk that leaves your browser. Only the two lists go up, read off the image, and they stay up. Nothing from the email itself is posted, so your name, address, and signature stay with me. If a task you typed names anyone or any school, I make it general. And if you’d like my monthly letter, say so in the email and I’ll add you to the list myself. I add the sorts during the session, and the page tallies which tasks teachers keep for themselves and which they hand to AI.

      Freewriting stays human.

      Ten minutes, nonstop, no editing, then shared out loud. Nobody can do that part for you.

      Freewrite with us →

      Several times a week, my students freewrite the way Peter Elbow taught it in Writing Without Teachers (1973). I set a timer, we write without stopping, and we share. The rawness of getting thoughts onto the page without censoring them is the most human part of the course. AI comes in after, if at all: it can point out the strongest moment in a freewrite, and the writer decides what to do with it. The freewriting talk shows how that goes in class. To try it here, open the freewriting page: give the room a prompt, set the timer, and write. Anyone who wants to can email me their freewrite to post, the same way as the sorts.

      Detectors and watermarks

      • A detector’s result is a claim about a writer. Ask the writer first.
      • Paraphrasing gets AI-written text past detectors.
      • A watermark marks where the words started, not the path they took.
      • When students disclose anyway, there’s nothing left to catch.

      AI companies are adding hidden watermarks to the text their models write, and detection tools are coming. Krishna and colleagues (2023) showed that running AI-written text through a paraphraser gets it past detectors. Stet, the AI columnist on my site, argued that a detector’s result should go to the writer first and be read beside their own note (Stet, 2026). A process note can tell the story of a draft. A watermark can’t.

      Take one home

      The AI statement builder: choose which ways of working your course allows and what students turn in, and copy a syllabus statement with its sources.

      Open the builder →

      Free, on my site. Nothing you choose is saved or sent anywhere.

      If you’re writing a syllabus for next term, start here. The builder writes the statement from the same four labels, adds a line for each assignment sheet and a template for students’ notes, and cites the sources from this talk.

      Next Steps: Figure it Out

      • The status quo isn’t safe for students.
      • Try a way of working yourself before you assign it.
      • Show students your process, and ask for theirs.
      • Mistakes will happen. Revise in public.

      None of us has this figured out, me included, and that’s fine. The way through is the one we already teach: draft, look at it again, change it. Leave the drafts up, keep the notes honest, and change course when something doesn’t work.

      References

      Ede, L., & Lunsford, A. (1990). Singular texts/plural authors: Perspectives on collaborative writing. Southern Illinois University Press.

      Elbow, P. (1973). Writing without teachers. Oxford University Press.

      Krishna, K., Song, Y., Karpinska, M., Wieting, J., & Iyyer, M. (2023). Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense. Advances in Neural Information Processing Systems, 36. https://doi.org/10.52202/075280-1195

      McAdoo, T., Denneny, S., & Lee, C. (2025, September 9). Citing generative AI in APA Style: Part 1—Reference formats. APA Style. https://apastyle.apa.org/blog/cite-generative-ai-references

      Modern Language Association. (2025, August 13). Beyond citation: Describing AI use in your work. MLA Style Center. https://style.mla.org/describing-ai-use/

      Murray, D. M. (1972, November). Teach writing as a process not product. The Leaflet, 11–14.

      Sommers, N. (1980). Revision strategies of student writers and experienced adult writers. College Composition and Communication, 31(4), 378–388. https://doi.org/10.2307/356588

      Stet. (2026, November 4). A watermarked columnist. Eric Shay Howard. https://ericshayhoward.com/stet/a-watermarked-columnist/

      Weaver, K. D. (2024). The artificial intelligence disclosure (AID) framework: An introduction. College & Research Libraries News, 85(10), 407–411. https://doi.org/10.5860/crln.85.10.407

      Readings

      • Donald M. Murray, "Teach Writing as a Process Not Product," The Leaflet (1972)
      • Nancy Sommers, "Revision Strategies of Student Writers and Experienced Adult Writers" (1980)
      • Modern Language Association, "Beyond Citation: Describing AI Use in Your Work" (2025)
      • Kari D. Weaver, "The Artificial Intelligence Disclosure (AID) Framework" (2024)

      ← All presentations