The new artificial intelligence bot ChatGPT has produced confident punditry about Japan’s economy, a Dr. Seuss-style retelling of London history and an imagined play starring a New York Times journalist debating free speech with a Silicon Valley techie.
Examples have cropped up all over the place in recent weeks. So, inevitably, the kids were going to get some ideas when homework came due.
San Francisco middle school teacher Andrew Bader shared on Twitter last week that an eighth grader submitted an AI-generated book report on Summer of the Mariposas by Mexican American author Guadalupe Garcia McCall—and got a passing grade.
That same day, Bader said, the school warned staff to look out for assignments made by OpenAI, the San Francisco company behind ChatGPT. Using its detection software, an educator found students “dabbling with AI” to complete their homework, according to an email obtained by The Standard.
“Supposedly, it’s a lot harder to detect with plagiarism software,” Bader said. “Technology’s always one step ahead.”
While some educators may embrace machine learning, a group of teachers at Oceana High School in Pacifica have their reservations. After seeing ChatGPT in the news earlier this month, they came up with some general messaging to share with students until administrators put out something official. And they cautioned fellow teachers to look out for AI-authored homework by using the detection software.
The warning made its way through word of mouth to San Francisco middle school where Bader works.
“Plagiarism is always an issue,” said Adam Weiss, a librarian at Oceana High who helped his colleagues hammer out the messaging about AI-made homework. “This just seems like more of the same, just the next sort of development. It’s easier than ever, but also the kind of thing we can figure out pretty quickly. It’s a lot easier to discover and recognize than kids give us credit for.”
Rather than wait for the San Francisco Unified School District to develop a policy or even block the AI software on school devices, Bader may start requiring handwritten essays or multimedia assignments that students can’t copy-and-paste from AI.
Teachers can often detect plagiarism when passages or entire assignments aren’t in the student’s voice, but Weiss said a lot of it is unintentional and should be a teaching moment instead of something to penalize.
But ChatGPT could have its upsides—as it does in classrooms like Owen Peery’s. The game design teacher at Balboa High School said he noticed a student using the program to spit out code for some classwork.
Peery said it dawned on him that the bot couldn’t do the work for students—they still needed to understand coding to know what the bot does or which sections to copy and paste for the games they were building. For students stuck on a blank page, Peery said the software could serve as a baseline to get started.
“It just made me think that there’s a lot of ways we could use this,” he said. “I know they’re smarter than a bot and this artificial intelligence stuff. Maybe for a high school English teacher this presents more problems but for me […] I have some ideas.”
In a group chat with Peery’s fellow teachers, ChatGPT has become a major topic of conversation. Technology-focused instructors like him are excited, while liberal arts teachers tend to be more nervous about the implications of AI in the hands of students.
Peery said a friend who’s an English teacher told him about watching in horror as her son used AI while working on his college applications—though not for the essays themselves, but to help formulate a starting point.
“Seeing how he was planning to use it made her reconsider,” Peery said. “Her biggest concern is that kids and adults struggling to write are working through their thought process. Long term, what might this do to our thinking abilities?”