As Brookline High School English teacher Robert Primmer graded his juniors’ papers last spring, he began to spot a peculiar trend.
“There would be these quotes from the text that were supporting some idea, and the idea was in the book, but the quote wasn’t,” Primmer said. “It wasn’t what the writer had written in any way.”
These made-up quotes, Primmer concluded, were likely authored by ChatGPT, an artificial intelligence (AI) chatbot that has become ubiquitous and raised major questions for educators since its release last November.
Primmer was no AI luddite. He had read about the tool and even experimented with using it to craft a midterm exam. But ChatGPT’s prose showing up in his students’ work felt like a new frontier.
“It’s going to fundamentally change everything,” Primmer recalled thinking.
This week, the high school hurtles into its first full year amid this new technological reality. Administrators and teachers are grappling with how to approach a rapidly-improving tool, and have reached little consensus about what level of usage to permit or about how to adapt their courses in response.
The “Wild West”
Within a few months of ChatGPT’s release, students began to use the tool widely for both the clearly impermissible, such as asking the chatbot to write an essay, and the more ambiguous, such as asking it to prepare notes or answer subject-specific questions, according to junior Dylan Brody.
“It’s the Wild West,” Brody said. “People are doing anything and everything.”
During peer editing activities, Brody said he came across sections of his classmates’ essays that were clearly written by ChatGPT.
“It was very obvious to tell what was just copy and pasted off ChatGPT and even what was tweaked a little bit,” he said.
Dean of Students Summer Williams wrote in an email that her office saw an “uptick in academic dishonesty by way of AI supported mechanisms” last spring. Two teachers interviewed for this story reported catching students submitting AI-generated writing.
Teachers said they are increasingly relying on an AI detection tool recently released by Turnitin, an academic integrity software company. Primmer, who initially caught the made-up quotes in his student’s essays on his own, said those papers received an AI-likelihood score of 85 or 95 percent on Turnitin.
But Turnitin and other tools like it are not perfect. The company has acknowledged that its AI detector can give false positives, according to the Washington Post. Some researchers worry that AI detection software might, for example, be more likely to flag the work of students who speak English as a second language.
A decentralized approach to ChatGPT policy
Catching AI-generated text in papers may be the simplest of the challenges ChatGPT poses.
In interviews, teachers differed on how they plan to address a vast gray area of usage, such as turning to ChatGPT for help outlining or brainstorming. They also questioned whether it would even be possible to police behavior like that.
Last year, the high school’s legislature amended its academic honesty statement, requiring for the first time that each department craft and maintain its own policy. (That change was made only partly in response to ChatGPT, according to Britt Stevens, a curriculum coordinator and legislature member.)
In their policies, departments have taken varying approaches to handling ChatGPT.
The social studies department’s policy bans all use of the chatbot, whereas the career and technology education department will not ban any usage, but instead require that students disclose when they use AI tools for assignments.
“We don’t believe in the abstinence approach to emerging tech,” said Stevens, who leads the career and technology education department. “That’s not who we are. So for us, academic integrity means being honest when you use tools that are not the tools that we assign you.”
The English and math departments, meanwhile, have not settled on policies and will spend the year considering various approaches, the curriculum coordinators for those departments said.
“I think [teachers’] initial gut reaction is, ‘This is cheating and we need to stop it,’” English department curriculum coordinator John Andrews said. “I think we need to rethink that a little bit because I don’t think it’s going to go away.”
The lack of a uniform policy means that students even within the same department may be held to different rules.
English teacher Elon Fischer said he will allow his students to use ChatGPT to summarize complex articles from secondary sources and is ambivalent about whether they can use it to help outline written work.
Primmer, meanwhile, is less inclined to allow these uses of ChatGPT. He has added a section to his syllabus which addresses his philosophy on the tool, though he won’t list where he stands on each potential use case.
Some teachers and administrators worry the range of policies may be challenging for students to keep straight.
“It’s really hard when you go to seven different classes and the rules are different in all seven of them,” Stevens said. “But getting us all to agree to the same thing when we do very different things is [also] hard.”
A new way of teaching
Whatever their policies, teachers agree they will need to restructure courses in a way that limits students’ temptation or ability to use ChatGPT in the first place.
Many teachers plan to convert papers into projects or in-class essays, cutting down on out-of-class writing.
“You’re going to see people handing out blue books and test books and essay papers,” predicted Brookline High’s Assistant Head of School Hal Mason. “I’m certain there’s going to be a lot more of that this year.”
Primmer anticipates spending more time than he typically does at the beginning of the year on in-class writing so that he can learn students’ writing styles. He also expects to place an increased emphasis on in-class discussion. He does not plan to completely eliminate out-of-class writing, but he will limit it.
“I will miss reading marvelous papers,” Primmer acknowledged.
Fischer plans to take a slightly different approach: He will continue to assign writing assignments as he has in years past, but he will test his prompts on ChatGPT first.
“I’ll enter it in and see, ‘Is ChatGPT capable of answering this question or not?’” Fischer said. “And if it is, it makes me think maybe this isn’t such a good question.”
Brody, the high school junior, believes the school is taking a short-sighted approach.
“What needs to happen is less running away and more embracing,” he said. “Let’s have teaching about how to use ChatGPT and let’s make sure every student is confident with using ChatGPT.”
Mason said administrators have not yet contemplated in a centralized way how to train teachers on appropriate ways to incorporate this new technology into their pedagogy, although he anticipates that will be something curriculum coordinators will consider this year.
For now, most teachers interviewed appear focused on preservation and adaptation rather than innovation.
“I’m struggling to be optimistic,” Primmer said. But, “I still have to find some practical thing to do. You still have to figure out what English class looks like.”