Scaffolding Student Writing in the Age of AI

How did writing with this formula help you, and how did it get in the way?

---

The Important Work is a space for writing instructors at all levels—high school, college, and beyond—to share reflections about teaching writing in the era of generative AI.

This week’s post is by Jennifer Trainor and Todd Walker. Jennifer teaches at San Francisco State University, and is a faculty director at SFSU's Center for Teaching Excellence. You can read her work on GenAI here. Todd teaches writing and literature courses at Clark College and Portland Community College. He has co-authored articles on assessment and writing pedagogy.

If you’re interested in sharing a reflection for The Important Work, you can find all the information here. Your reflection does not have to be about using AI in the classroom; we are interested in any way that you’re thinking about the important work these days.—Jane Rosenzweig

---

(Mostly) gone are the early ChatGPT days of sing-songy synthetic essays. Instead, in our classes, we’re seeing hybrid drafts and hearing from students that they “use AI for brainstorming” or “to get ideas” or to help put ideas in the “right format.” We also encounter hybrid drafts — students’ writing intermixed with authoritative clichés, soothing summaries, or bland topic overviews.

Many of our students are first-generation, minoritized, and working class. Academic writing does not always come easily to them. As new—and sometimes alarming—as our students’ use of AI may be, we have begun to think about AI tools as scaffolds. That is, AI outputs may be similar to formulas and templates (from the five-paragraph essay form to the quote sandwich) that can stymie but also sometimes support novice writers.

We use metacognition to help students question formulas rather than relying on them uncritically. We believe students need a similar approach when it comes to AI. AI outputs—from AI-generated outlines to explanations of ideas to rephrasing and editing with tools like Grammarly’s AI Writing Assistance—can be understood as scaffolds: prone to reification, reductivism, and over-reliance, yet useful for managing complexity and difficulty. If the key to helping students gain rhetorical agency over formulas is metacognition, perhaps a metacognitive pedagogy can help students use AI to support rather than supplant their own thinking, ideas, and voice.

Photo by Scott Webb on Unsplash

Our goals here align with the work of Chris Mays, who argues that formulas can be useful but only if we recognize their limitations. Formulas offer “temporary stabilities.” They allow students to engage with smaller slices of larger, complex systems. The problem, as Mays notes, is that sometimes novice writers mistake the scaffold or formula for the whole. With that in mind, we’ve been working on ways to help students develop a metacognitive understanding of the ways that GenAI can support their growth as writers, rather than write for them, functioning as a scaffold rather than a final draft.

We have not yet fully integrated AI into our assignments. Jennifer has co-created assignment guidelines with students over the past few semesters; Todd follows department guidelines that prohibit unauthorized use. But we have both led class discussions about GenAI and how we might use it ethically. Drawing on this preliminary work, we want to share our ideas for incorporating metacognitive GenAI use in the fall.

Our question: can we use metacognition to foster awareness that formulas and AI tools are scaffolds—temporary stabilities, in Mays’ words—that writers use to manage complexity? Can we create opportunities for metacognition that move students from reliance on to experimentation with stabilities, including AI? We imagine students working at least momentarily in between stabilities, untethered and able to see stabilities from a birds-eye view.

Leaning into the Overlap Between Writing Formulas and GenAI

We typically begin the semester by asking students to reflect on the formulas they’ve learned in the past and to consider how those shape their writing. We now ask those same reflection questions about AI outputs. Student responses, as Jennifer has written elsewhere, are revealing:

• The kind of writing that Chat does, it’s what our teachers try to get us to do.

• It’s like five-paragraph essays, and perfect paragraph[s] that don’t have any personality, which we were taught in high school.

• It does what school has trained us to do. Like write a perfectly formatted essay that is based on some random people’s ideas.

We routinely ask students to reflect on formulas for writing. This fall, we plan to include metacognitive questions about GenAI: Did AI help you express yourself or get in the way? How has AI improved your writing in the past? Fostered your confidence? Facilitated your ideas? What problems, if any, did you encounter as you used AI? How did it shape your writing? To what extent did it help you navigate the hard parts of writing?

Turning Pre-reading Strategies into Pre-prompting Habits

Our students often struggle to find a way into difficult academic texts. Students have told us that AI outputs are similarly difficult for them. “It overrides me,” as one of Jennifer’s students said. Pre-reading questions can help students wrestle with difficult texts and may be adapted to promote metacognitive AI use. In other words, pre-prompting questions may give students more agency over AI output. What kind of help are you looking for with AI? Brainstorming support? Help with transitions? Counter-arguments to juice your thinking? An illustration for a point you are struggling to put into words? What do you already know about the topic? What kind of AI responses do you predict?

After they read AI output, we can ask students to reflect: Which outputs were helpful and how? Where did the AI fall short and why?

Embracing AI as a Sentence-Booster

Although AI often uses a robotic style, students tell us that some AI-generated sentences, even the sing-songy ones, can catalyze their writing, inspiring new ways to express their ideas. As Donna Gorrell long-ago noted, imitating sentence models is an ancient, creativity-enabling practice. Our students have reported that copy/pasting some AI sentences into their draft gives their writing “a boost,” much in the way sentence templates like those in They Say / I Say can support novice writers. To ensure that such moments are scaffolds rather than crutches, we will need to ask students to reflect: What does AI do with sentences that you typically don’t do? Which AI sentences surprised or inspired you? What new words did you learn from AI? Which of your own sentences were modeled on AI? Which AI sentences did you rewrite? How did you rewrite them?

Empowering Students to Think Rhetorically About AI

We regularly ask questions such as “how did writing with this formula help you, and how did it get in the way?” Students report that formulas are constraining, but, as Mays suggests, they also find them useful at times. For example, students notice that the quote sandwich formula can be a useful container for their ideas or can help them analyze a quote. Our goal has long been to get students to notice the formula as a scaffold that can be both enabling and constraining. This semester, some of our students reported that they regularly use AI to generate ideas or create an essay outline. When they do, they may use some of the phrasing or content of AI-generated topic sentences, but put in their own examples, or they may use the AI-generated output as a model for structuring their own draft. We agree with students who say, this is a good use of AI, partly because this use of AI closely mimics a scaffolded assignment that provides similar kinds of structure for students. But using AI in this way is problematic if we don’t also help students develop the agency that metacognition builds: How did you expand the AI output? Which parts of the output did you abandon, and why? Which parts of the output inspired you or supported your drafting?

Like many teachers, we have developed activities that ask students to compare student writing to AI outputs, but we believe we need to do more than merely show AI as robotic or untrustworthy. Instead, we want students to reflect on its potential use as a scaffold: What moves did the AI make that you didn’t? What moves did you make that the AI didn’t? What do you think accounts for these differences? Where is your writing stronger? What ideas / sentences did the AI give you and how might you use them in your draft? Where did the AI help you clarify something?

Formulas ground students’ texts in ways that can easily be mistaken for writing. Students may integrate a quotation correctly, following a template on “quote sandwiches,” but fail to understand the relationship between the quotation and their rhetorical purpose. A metacognitive approach unsettles this grounding — we picture students floating momentarily unmoored, as what once seemed like a fixed formula becomes a more flexible heuristic.

We remain concerned that AI tools violate values we hold dear. Using metacognition helps recenter those values, by demoting AI to a temporary stability and recentering metacognition as the value of the class. Same as it ever was? As John Warner pointed out early on, the problems of AI-generated writing and formulaic writing are the same. The antidote to formulaic writing—metacognition—offers a way forward in the age of AI as well.