By Juliet Rotondo
Rye High School teacher Joel Edminster is changing the way he teaches AP English Literature, starting with the summer assignment students complete before the first day of classes in September.
In previous years, Edminster required students to read two novels and write essays responding to each. This fall, he is planning to replace the essays with in-class exams administered over the first few days of school.
The reason for the change?
To deter students from using artificial intelligence to write essays before they set foot in the classroom.
“It’s harsh, and I’m sure it will raise anxiety levels,” Edminster said. But he believes this change is necessary to combat the uptick in AI use he has observed over the last few years.
As large language models like Chat GPT and Gemini become increasingly advanced, students have been turning to AI to revise, rewrite, and even replace their own writing — forcing teachers to find ways to combat academic dishonesty.
Scanning essays for evidence of AI has become a routine part of the grading process for RHS English teachers.
When AI first became popular two years ago, English teacher Anthony Frabizzio used the plagiarism-detection platform Turnitin to check whether students had submitted papers written by chatbots. But as AI evolved, Turnitin became unreliable, flagging essays that students had written themselves.
Now, Frabizzio looks for other clues. Students’ Google Docs histories can reveal AI assistance; if a student quickly generated a polished essay instead of writing and revising over time, it’s likely they enlisted the help of AI.
Certain words and grammar choices can indicate AI-generated writing, but sometimes those are the product of a specific teaching style rather than AI.
“One of the characteristics of AI is to use participial and absolute phrases, which I teach in the composition class,” Edminster said. “That’s a mark of good writing. But AI always has an introductory participial phrase, and it’s almost a red flag for me now when I see it.”
English teachers describe the process of checking for AI use as miserable, frustrating, and disheartening — but unavoidable.
“I don’t want to read things written by a robot,” Frabizzio said. “I don’t want to give feedback to something you didn’t write.”
And when the time comes to confront a student about AI use — something teachers can suspect but not prove — it puts teachers “in an awkward position,” said Edminster, who will speak to students privately about their use of AI and the resulting consequences.
Officially, the penalty for cheating is a zero in the gradebook.
“I do my best to follow the English Department policy, but the English Department policy is in flux right now because of the advancements of AI,” Edminster said.
Added English teacher Roni Sarig: “The district is trying to get ahead of it and put some policies and understandings in place. But it’s still a confusing time in terms of how people can use it versus how people shouldn’t use it.”
The district has encouraged experimentation with AI, and some teachers see its potential as a productive tool.
Frabizzio described a class where he encouraged students to get feedback from AI on their essays by “asking really good, interesting questions about their writing.” Instead of asking AI to do the writing for them, students used conversations with AI to learn more about how to approach a writing assignment — as they would if getting feedback from peers or from Frabizzio himself.
A major drawback of using AI to write is that “learning to write is learning to think, and you’re cutting that out every time you jump straight to the final product,” Sarig said.
Edminster and Frabizzio agree — as do researchers at MIT’s Media Lab, who found in a recent study that those using ChatGPT to write essays show lower brain engagement than writers who use Google’s search engine or nothing at all.
The researchers warned that relying on AI for “immediate convenience” may come at the cost of “long-term brain development,” according to a Time magazine article on the MIT study. Psychiatrist Dr. Zishan Khan, told Time that he finds young students with developing brains who rely on large language models for their schoolwork are weakening the neural connections that help access information and memorize facts.
Administrators open to AI use seem to hope that if schools properly integrate AI into their curriculum, students might learn from a young age to use AI to supplement the thinking process rather than get around it. That is why researchers who have studied the impact of large language models on brain development believe it’s critical to raise their concerns about AI use in schools.
Nataliya Kosmyna, the main author of a paper on the study’s results, said she wanted to publish researchers’ findings about the use of large language models because she fears that “in six to eight months, there will be some policymaker who decides, ‘let’s do GPT kindergarten.’ Developing brains are at the highest risk.”
Back in Rye, Frabizzio describes using AI as a teaching assistant as “dicey” for these reasons and more. Large language models are owned by corporations with agendas, and their biases affect how the algorithm is developed and what information the chatbot chooses to provide.
If all high school students were to learn using the same large language model, Frabizzio suspects, “they would get a very narrow band of research out of it.” What happens when you funnel every student through AI? Will everyone sound the same and parrot the same interests? “That’s the ultimate disaster,” Frabizzio said.
If given the choice, Edminster would not use AI at all; he would assign essays to be written in class using pen and pencil.
“I’m open to new innovations, and I would feel terrible to make all my students handwrite everything,” Edminster said. “But if I were given the choice, I would choose to say ‘No.’”
