Artificial Intelligence and Writing: Four Things I Learned Listening to my High School Students
A guest post from Brett Vogelsinger
Because I am something of a skeptic when it comes to the utility of the tools of generative AI when it comes to learning to write, I think it’s important to get additional perspectives about how and when students are using these tools. In this case, my guest writer, Brett Vogelsinger has gone to another authority, the students themselves. This is what he’s learned.
Brett Vogelsinger is an English teacher at Central Bucks High School South in Pennsylvania with over two decades of experience teaching in middle and high school. He is the author Poetry Pauses: Teaching With Poems to Elevate Student Writing in All Genres, and his book on the use of generative AI in secondary writing instruction is coming soon from Corwin Literacy. He is the founder of Go Poems, a frequent contributor and webinar creator for Moving Writers, and a professional development presenter. Connect with him on LinkedIn, @theVogelman on Instagram, Threads, or X. His website is www.brettvogelsinger.com.
Artificial Intelligence and Writing: Four Things I Learned Listening to my High School Students
By Brett Vogelsinger
As a high school teacher with two decades of experience, I am old enough to know that no matter what approach I help teenaged writers take to their work, writing is not easy. It should not be easy. Like any work of art, when we sit down to transform a blank page or screen into something new, we undergo struggles to wrangle out something fresh, to clarify our own thinking, to convey concepts in the just-right way. We choose and we trim and we polish until we have something of sense and beauty.
And if we are being honest, not all high school writing arrives at beauty.
That is OK. They are learning. If we can take them beyond the formulaic, helping them develop a unique voice and reliable skills and knowledge that support writerly choices, we are doing our work as teachers well.
When generative AI tools became broadly available, the first thing that raised my hackles was the promise of ease. The new technology created an illusion of easy writing. It reinforces formulaic approaches, predicting what some hypothetical human might say about a topic rather than what the very real, developing human in front of me in class wants to say.
Then came a wave of enthusiasts and evangelists, promising that AI technology would “revolutionize” teaching and learning. Few of those voices, I noticed, had actual rosters of students to teach.
Since then, many of the high school teachers I look up to most have maintained a public silence on the topic, perhaps confiding their early takes in personal conversation, but nervous to lock in any public stance. They are biding their time, watching and learning, and I do not fault them for this. In teaching, patience often wins.
Others have decided to go on record as proudly banning AI in their classes until the educational, ethical, and environmental ramifications are clearer.
I respect all of these stances, especially coming from people doing the hands-on work of writing instruction with students, a not-easy task as students engage in a not-easy process.
I think another approach to AI is an option, listening to our students.
Students find AI fascinating in all the ways adults do. Some delight in speed of assignment completion it promises while others recoil from the dystopian replacement of creativity it threatens.
In the past year, I have listened to some students who have attempted to cheat with generative AI, some who, at my invitation, have tried tinkering with it in specific ways, some who have played with it freely on their own time with the understanding that conversations with me would be part of their process, and some who have banned themselves from using it, because, as one of my tenth-graders put it in a survey, “there’s no growing as a writer in having something else write it for you. I ‘grew as a writer’ [this semester] by writing it myself.”
But the biggest discoveries I have made through listening have nothing to do with AI. They have a lot to do with how my students learn, write, and perceive themselves as writers. They have a lot to do with personal ethics and family perspectives on school. Most exciting, students are talking more about what they value about their own voice becauseAI exists. They notice the hollow sound of AI-generated text and how unsatisfying it feels to read. They attend more to how enjoyable it is to read something human-crafted, even in a rough, rough draft.
Keep in mind, this is all happening in a classroom and a school system where students have seen their teachers model writing, where teachers use mentor texts and exemplars to teach, and where I ban the writing of five-paragraph essays in my class. Foundations matter!
One of the beautiful opportunities of high school is the face time we get to have with students. My school, like many throughout the country, has built in some time for intervention and enrichment during our day – we call it Lunch and Learn – and during this time I have enjoyed extended sit-down talks with students about their writing and how AI interacts with it, how it could be hurtful or helpful to their growth as writers and thinkers.
Here are four takeaways I have discovered:
1. Students care about how vocabulary and syntax affect their style. When kids feel their own writing is inferior to what AI can create, what they often mean was that the LLM used vocabulary and syntax to create an authoritative tone in the writing that the student was not able to muster yet. This emphasizes the importance of direct instruction in vocabulary and syntactic structures that help students to keep growing in the complexity of their sentences during middle school, high school, and college. Our English language is larger than most with endless ways to compose and arrange our words in ways that are almost musical in their quality. Embedding more of this into our instruction helps students to develop their own tone of authority. A conversation about an AI-generated draft can spark this development. We can ask: “Where do you think this sounds better than something you could write on this topic? Highlight three sentences. Let’s talk about what is going on in these sentences and how you can learn to write with the sentence structures that sounds so good to you.” And of course at the core of direct instruction on these tactics should be human-created mentor texts and mentor sentence from master writers and from you, their writing teacher. (Check out the YouTube channel Mini Moves for Writers to build skills in this practice!)
2. Students rarely get to talk about their writing process and what to do when they feel stuck. When a student in a high school class is stuck, that often manifests as avoidance in the classroom (i.e. chattiness with peers) or at home (i.e. the procrastinator’s slumped frame facing off with the midnight glow of a laptop screen). It is worth exploring whether LLMs can help students get “unstuck” and more forward with their own writing. Perhaps students can tease out a sentence from an AI draft that helps them find focus or build a little momentum or illuminate a new corner of a topic they would not have considered investigating. This involves caution and credit of course. Listening to students about their struggles and talking to students about this potential use lets us help students see beyond the use of AI tools as simple cheat mechanisms.
3. Students (most of them) do not feel attached to AI-generated language, even when it helps them in their process or creates an illusion of ease. When students share a first draft with me that I’m pretty darn sure is generated by a ChatGPT, my first reaction is, “I don’t really like how this sounds. I’d like you to try it again.” Since the students are not attached to the language of the draft – it was unethically developed in seconds and not born of any struggle – they are often quick to move on and agree to give it a second try. After seeing that more human draft, I will question whether the first was AI generated and ask, “What do you think tipped me off? What did you notice about the style of the writing?” Evaluating what AI creates helps students to differentiate between language that is clear and direct and writing that is crafted with meaning and purpose.
4. Inviting students to dabble in AI at particular stages in their writing process opens productive conversation. I hope that we can sometimes have rich revision activities in the high school classroom that pinpoint how to take bland writing from AI and go deeper and more nuanced into a topic in ways a machine cannot. Students may first try AI outside a teacher’s presence to “get ideas” for their writing. We could cringe at that, or we could listen to what worked about that, what felt false to them, and how they evaluated, changed, or abandoned ideas that AI generated as predictably human ideas on a topic. We will learn lots from this listening that will help guide us in wise next steps.
Reflecting on research work with college students on a later stage of their revision process, Patricia R. Taylor and Mark C. Marino concluded, “What was more valuable than the feedback [AI tools] gave to students were the conversations that arose . . . it opened a conversation about what we are seeking in feedback styles and feedback content” (2024, p. 6, para. 2). I have found this is true with high school students as well, as I experiment with integrating AI into my writing instruction. Sometimes the positive outcome is not the writing, but the conversation about writing that their experimentation opens.
In their document, Education Hazards of Generative AI, Paul Bruno and Benjamin Riley of Cognitive Resonance comment that “Leaders should not invest time and resources to incorporate AI in schools based on assumptions about what the future will bring. Nor should they drastically alter curricula to prepare students for an “AI world.” We simply do not know what such a world will look like or what it will require of future citizens” (2024 p. 11, para. 6). Listening to students in our ongoing explorations will help us be future ready in a different way. Whatever changes are ahead, conversation between humans will be the best antidote to diminishing or destroying the quality of student writing and the instruction we provide to support it.
The phrase “human-centered AI” has caught on recently, and nothing can be more human-centered than sitting down together with one of my students to talk about their writing process, their choices, the struggles they must encounter to produce something artful and wise.
That will never and should never be easy.
Works Cited
Marino, M. and Taylor, P. (5 August 2024). “On feedback from bots: Intelligence tests and teaching writing.” Journal of Applied Learning & Teaching, Vol. 7, No. 2. https://doi.org/10.37074/jalt.2024.7.2.22
Riley, B., & Bruno, P. (2024). Education hazards of generative AI. Cognitive Resonance. https://www.cognitiveresonance.net/EducationHazardsofGenerativeAI.pdf
Some of the most profound insights into the real ways AI will change education come from teachers already exploring this frontier with their students. Thanks for sharing these rich lessons!
Thank you for sharing this, Brett (and John). I'd like to pick up on the point about being stuck, because I agree that this is a place where students often don't get enough - or the right kind - of support.
To provide that support, it's important to think and talk with our students about why writers get stuck. In doing so, it's important to distinguish between different kinds and causes of stuck-ness. Different kinds of stuck may call for different strategies for getting un-stuck.
For example, we might be stuck because:
1. We're alienated from the work. Getting unstuck is a matter of finding our way to a more meaningful kind of work and/or a more meaningful relationship with it.
2. The resources that we need to do the work have been depleted by (mostly) external factors. Getting unstuck is a matter of dealing with adversity and adapting to our environment.
3. The resources that we need to do the work have been depleted because we've been approaching our work in a way that grinds us down. Getting unstuck is a matter of developing a more sustainable practice.
4. We don't have an effective strategy for doing the work (i.e., we don't know where to start or what to do next). Getting unstuck is a matter of developing a more effective process.
5. We're encountering intellectual difficulties that are intrinsic to the task. In this case, we may want to get not unstuck but "stuck in."
Of course, we often get stuck for more than one reason at once, but even when that's the case, it can be helpful to break our stuck-ness down into its component parts and try to understand how they interact.
The thing that I think is crucial to understand is that getting stuck in any of the ways outlined above can lead to really valuable learning moments. The goal, I think, should be to create an environment where students can recognize the potential value of getting stuck and work through it in a constructive way, rather than an environment where getting stuck is understood simply as an obstacle to delivering an end product.