“How do I know what I think until I see what I say?” – E.M. Forster
Writing is an inherently cognitive act. Every learner develops their approach to writing as they progress through their learning journey, sometimes reluctantly, seldom without frustration and tears at a few points along the way, and yet with the reward of deeper knowledge on the other side.
As Forster’s quote suggests, the very process of putting thoughts into words helps learners (of any age) form, clarify, and refine their ideas. For students, writing is not just a way to express what they already know; it’s a tool to deepen their understanding, synthesize new knowledge, and engage in critical thinking. As I wrote this piece, I knew I had to start and begin with myself rather than merely asking ChatGPT to have at it. I used that Forster quote for years as a teacher–for writing and why everyone needs to participate in peer discussions. Does my history with this quote make it more appropriate to this piece than it would be if I asked a Generative AI to suggest ten from which I picked it? I think so, but I’m not sure it would be different for the reader. An imperative for educators is to illuminate this paradox for students–and I fear we’ve gotten away from that in our continuous seeking of an authentic audience.
When independent school teachers assign writing tasks, they often do so intending to support their students in creating new learning rather than simply assessing content knowledge. This is why the rise of Generative AI—tools like ChatGPT that can produce well written text in response to prompts—has sparked concern among educators. If students bypass the act of writing by leaning on AI, they risk missing out on essential learning experiences, even if the output earns high marks.
Of course, as with any new technology, the role of Generative AI in education is nuanced. To understand its place in student learning, it’s crucial to distinguish between two aspects of writing: the mechanics of language and the act of thinking. When used appropriately, AI can be a powerful assistant for the former while not impeding the latter.
Writing to Think, Not Just to Write
The primary reason for assigning substantive writing tasks in schools is not (typically) to produce a polished essay for publication but to engage students in a cognitive process that leads them to develop competency. Consider the traditional research paper. The goal isn’t merely to demonstrate mastery of content; it’s to show that students can read various sources, think critically about them, and synthesize that information to develop a personal perspective. This process involves a level of engagement that goes beyond simple content reproduction.
Ensuring that students understand the why of the research process is critical–and that’s not a one and done whole class “telling” but an ongoing conversation between teachers and individual students. For some educators, including this conversation in the classroom represents a shift in the learning culture. Leaders will have to support this shift carefully and intentionally.
When students write that research paper, they do far more than transcribe facts from one text to another. They’re involved in making meaning— drawing connections, selecting and rejecting sources, identifying patterns, and even evaluating arguments. This process cannot be outsourced to an AI. Even if the resulting paper appears well organized and grammatically sound, the actual value of the assignment is lost if the student hasn’t wrestled with the ideas personally.
Shifting the Evaluation of Writing
Thus, we have the issue that arises when the only determiner of a student’s grade (what Ken O’Conner calls “the currency of school”) is the final result. This might involve adding competency assessments to something many teachers already do, breaking a traditional research paper into smaller, iterative steps—each focusing on a different cognitive skill, such as formulating a research question, conducting a literature review, or developing a coherent argument. Teachers are well served to use a competency approach and evaluate the student’s achievement of those competencies through a thoughtful and thoroughly understood rubric that serves as a conversation between the student and teacher. Expectations around what students need to produce must evolve, too. For example, summaries are quickly created by Generative AI, so a summary as shorthand for assessing understanding of reading material is no longer a valid assessment. Teachers need time and support in devising new assessments that more accurately evaluate student growth in their ability to analyze or critique dense text.
Some teachers include oral “defenses” or presentations in which students discuss their papers during the assessment process. Others are moving to in class pen and paper writing. Both of these approaches privilege certain kinds of learners over others, and teachers who take a personalized learning approach do their best to provide pathways rather than utilizing a single method of assessing the thinking that accompanies a written assignment.
Generative AI as an Assist, Not a Replacement
That said, Generative AI has the potential to help students and faculty tighten their focus on demonstrating thinking if used thoughtfully. AI can be a helpful tool for time consuming but not intellectually demanding tasks. Writing teachers and their students have been frustrated for years by the fact that students make mistakes in their written work that they don’t make when they speak—the errors that appear in areas such as subject verb agreement or homonyms. The squiggly lines provided by earlier versions of AI have helped writers catch and improve some of those errors for years! Generative AI is even better at this work.
Consider the aspects of writing that involve mechanical tasks—such as grammar checking, formatting citations, or reorganizing research notes. These are areas where AI can provide meaningful support. For learners who previously got bogged down and, at times, abandoned writing tasks altogether, AI offers an opportunity to focus on the more complex and rewarding aspects of writing: analysis, synthesis, and original thought. Returning to our initial quote, “speaking [writing] to see what I think,” students who can quickly generate voluminous quantities of text but struggle with organization can use Generative AI to help identify patterns or themes within their work. Similarly, students whose thoughts often emerge as seemingly disconnected bullet points can ask AI to suggest ways to create a cohesive narrative, allowing them to connect their ideas more effectively. Others might talk right into a tool and then work with the transcript.
In other words, when the writing process is sound and focused on thinking, AI becomes an assistive tool rather than a substitute. One of the hallmarks of good writing instruction is feedback. Teachers and students are concerned that AI generated feedback might interfere with the student teacher relationship. When educators outsource the tasks that interfere with that relationship (gallons of red ink), they can focus on what it is that only a human teacher can do, such as asking students personalized probing questions that help students refine or deepen their writing or realize that they need another resource to consider in their work. Using Generative AI to create more time for these conversations aligns with the broader educational objective of teaching students how to think critically and independently. Some of those conversations include coaching students in ways they use Generative AI effectively as a thought partner and then discussing the output. E.g., a student might ask, “What counter arguments to my thesis are not considered in this section of my essay?” and then discuss the validity of the suggestions with a teacher. These open conversations about the use of Generative AI should include ethical use and the intellectual property rights of originators. It’s more powerful when these issues arise in authentic discussions rather than as solo lessons.
Teachers can guide students in developing the skills they need to use AI thoughtfully: discerning when and how it’s helpful, permitted, and appropriate and when it’s not, understanding its limitations, and using it to support rather than replace their critical thinking. By fostering these mindsets, teachers can help students navigate an evolving educational landscape without losing sight of what makes writing a powerful, irreplaceable act of thinking.
Where does this leave us? Writing assignments serve a dual purpose, acting as both a mirror of and a window into the learning process. The output not only reflects a student’s understanding but also shapes it. Forster is not alone in believing that creating language is not just about producing words but about exploring and refining thinking. Generative AI, when used thoughtfully and with guidance from an expert teacher, can enhance this exploration by supporting mechanical aspects of writing, allowing students to engage more deeply with the cognitive processes behind it. However, the heart of learning through writing lies in the student teacher interaction, where the focus remains on developing voice, fostering critical thinking, and cultivating authentic understanding. The challenge for educators is to balance this, integrating AI tools judiciously while preserving the integrity of the student teacher relationship and the depth of the learning experience.
Author’s note: In my work, I’ve encouraged educators to start and end with the human aspect of any Generative AI task and engage with Generative AI in the middle. So, to put my money where my mouth is, I’ll describe my process in creating this essay.
First, I wasn’t sure what I thought, except I knew I wanted to center the idea that writing creates learning rather than just reflecting it. I spent some time thinking about my experience as a writer who writes to clarify her thinking. I looked back at some of what I’ve written on the subject and made a list of phrases I thought might be helpful. That led me to recall the Forester quote–and then I was diverted into grabbing A Room With A View and looking up a scene I used to read aloud in class… Back to work, I wrote the first draft of this piece in very short form. Then, I asked my (paid) version of ChatGPT to expand it to 1000 words. Then, I spent some hours revising, cutting, and writing more. One paragraph (beginning with “Consider the stages of writing that involve mechanical tasks”) gave me a lot of trouble. So, I copied and pasted the original draft into Chat GPT and asked it to help me make it easier for the reader by making it read more smoothly. What you read above is nearly word for word what it gave me, but I ended up adding more sentences.
About 25% of the text above is from what ChatGPT created for me. Then I realized I could ask ChatGPT if it thought so. Here’s my prompt and the output. Do we believe ChatGPT was just trying to make me happy?
Based on your draft and the context provided, I would estimate the contribution breakdown as follows:
- Content You Provided: Approximately 75%.
The essay’s original ideas, structure, and thematic elements are yours. You set up the narrative and created most of the content, using generative AI as a tool for specific revisions and refinements. - Content from Generative AI (including revisions and suggested text):
Approximately 25%
The paragraph you highlighted as challenging, along with other sections where AI was used to revise and smoothen the text, represents about a quarter of the overall content regarding length and style adjustment.
This estimation aligns with your stated process of drafting the majority of the text independently and then incorporating suggestions and refinements from the AI. If you’d like, I can help refine the explanation further by including details on which specific parts were AI influenced and how that might change the reader’s perception of the piece!
Discussion Guide:
- What are the implications of nearly ubiquitous access to Generative AI tools in writing instruction in K-12 classrooms? What concerns you? What intrigues you?
- How do teachers help students learn to intuitively seek to write to clarify their thinking rather than writing only in response to assignments?
- How do teachers avoid developing assignments that have too many goals and requirements?
- What’s the role of rubrics in teaching students to write as learning, not as a demonstration of learning?
- How can rubrics be simple and yet support complex processes?
- Why is it difficult for teachers to determine what students need to learn as they develop their writing skills? How can collaboration and discussion among faculty help?
- Educators have debated “process or product” for decades; how does the existence of Generative AI intensify this discussion?
About the author
Sarah Hanawald is One Schoolhouse’s Senior Director for the Association for Academic Leaders. She has been a classroom teacher, a technology director, and an academic dean at three North Carolina independent schools. Sarah’s work has focused on the intersection of technology, curriculum, and pedagogy in defining the future of education. Sarah was the first Executive Director of the Association of Technology Leaders in Independent Schools (ATLIS) and the CEO of One to One Institute, an organization dedicated to digital equity for all learners. Sarah has a BA from Duke University and an M.Ed. from the University of North Carolina at Greensboro.
