The poem was perfectly passable–that was the problem. The words were there, arranged with technical precision, but they seemed to be gesturing at emotion rather than containing it. When the 10th graders in Cam Kaplan’s English class read it aloud, something felt hollow.
No one had written this poem. Or, rather, everyone had: every poet, every song, every lyric, every greeting card. ChatGPT had simply predicted what came next.
Historically, in the 10th-grade poetry unit, Kaplan has his students go through multiple revisions of their poems. The issue, he said, is that it can be “a little bit intimidating. I think it can be hard to revise your own work.” So, Kaplan had students use ChatGPT to “generate first drafts that were cliché on purpose and then revise them.”
ChatGPT is essentially a glorified prediction engine, using deep learning algorithms to predict the most likely sequence of words based on a user’s prompt and its vast training data. “It helps students see poetry that is maybe stylistically capable but lacks originality [and] feels kind of mechanical. It doesn’t feel emotionally moving, so that’s helpful,” he said.
“If their homework is to go home and write a poem, I know that they can generate it on ChatGPT,” Kaplan said. “But I also know that that poem is not going to be very good, and they’re going to spend as much time workshopping it to create that novelty and uniqueness as they would if they had written [an] okay first draft themselves.”
Is the editing process, then, what makes something one’s own work? Or is the lesson here that poetry’s novelty must be added by a human hand? Similar questions can be applied to most components of writing. “Your ability to evaluate what text is actually doing is starting to seem more and more valuable,” Kaplan said, in contrast to merely producing text.
This raises the uncomfortable concern: What if thinking is a muscle? What if we’ve just invented an algorithm to do our lifting for us?
Spanish teacher Jeff Mayfield can’t shake this image: “If you give a kid a calculator too early, they never learn addition, they never learn arithmetic. So, using AI, I just figure they might lose their own creativity.”
Maybe the answer is more unsettling than that. Perhaps it’s not just about creativity, but rather about the particular kind of agency that comes from wrestling with blankness until you discover what you think. What happens to that moment when the answer is always one prompt away?
AI demands a reevaluation of what students should take from classrooms and how teachers should teach them. A good balance, Head of Technology Jamie Sullivan said, is AI helping “with practical things…the day to day, creating rubrics or thinking about how to rework a lesson in a better way. [It just gives] you ideas. You don’t take them all, any more than you would take something you Googled and pasted somewhere else.”
Castilleja’s learning management system has also incorporated usage of AI, specifically to assist teachers. Mayfield says that Toddle has a feature which will “look at what [teachers] created as an assessment and give [them] feedback on.” However, he also said that “it’ll take a while before I can really assess whether it is actually more effective or not.”
Most teachers recognize the potential of AI in the classroom, especially as a tool for accessibility, differentiated instruction and personalized feedback. But many also insist that schools must proceed with caution, ensuring that technology enhances rather than replaces the human elements of teaching.
As Castilleja continues to experiment with these new tools, educators hope their perspective will remain central to decisions about how students learn in an AI-driven era. After all, they’re the ones guiding students through a newly redefined landscape. As Sullivan said, “Our goal is to empower young women… My job is to see how technology can support that and prepare students for the future.”
Empowerment has always meant the same thing: agency. The ability to be the author of your own thinking and the difference between having a tool serve you and serving the tool.
The technology will keep improving. Every student will face moments when the easy path is right there. But that “generate” button will not produce anything special; to create a truly memorable work, one cannot rely on what has already been created.
Maybe that’s the most important thing education can offer at this moment—not rules or restrictions, but discernment—the capacity to see the difference between competent and meaningful. In a world flooded with answers, it is crucial to understand why arriving at one’s own conclusion still matters.
