The Berkeley High Jacket


Newsletter

The best of the Jacket, delivered to your inbox.

News Print
October 4, 2024 Login
Features

BHS teachers rethink instruction amid rise in ChatGPT use

Robert Gellner
By Sage Feldman, April 21st, 2023

The current iteration of AI chatbot Chat Generative Pre-trained Transformer (Chat-GPT), GPT-4, has quickly developed into something difficult to ignore in everyday life. AI technologies like GPT-4 have made pernicious inroads into areas as varied as art, coding, and possibly education.

When Academic Choice (AC) English teacher Yoshi Salaverry asked the AI to write him an essay on the theme of unrequited love in “The Great Gatsby,” which he has read and discussed many times, he was impressed with the resulting analysis. “It was relatively effective,” said Salaverry, who believes that GPT-4 is very good at answering broader prompts like his. “I’ve had conversations about this with some of my colleagues, and what some of them have pointed out is that if you have … specific (or) narrow … requirements … it’s very hard for Chat-GPT to navigate its way through all of those very specific requisites,” he said. In addition to moving to more in-class, single-day assessments, Salaverry will consider reframing his essay questions to be far more specific.

Brooke McKinney, also an AC English teacher, views the emergence of the AI as a wake-up call for teachers to take a closer look at the prompts they offer to students. “Chat-GPT is really calling out the formulaic nature of a lot of essay writing,” she said. “It gives us the opportunity to rethink … how students can more creatively respond to something.” 

In her opinion, these essays were already close to being formulaic and pseudo-computarized. GPT-4 takes that a step further.   PAGE 11

McKinney was asked by the Jacket to read and grade four sample essays, each responding to the same reading and writing prompts. Two of the samples were written by students, one by GPT-4 that had been given the same prompt as the students, and one by GPT-4 with a prompt revised for a better response from the engine. The two AI responses were scored one letter grade higher than the student responses by McKinney, who did not know who wrote which essay. However, she was also able to correctly tell apart the computer-generated and the human-authored responses.

Where Salaverry sees potential in GPT-4 is not as a full-on ghost writer but as a revisor. “I don't think anybody is going to level up from being a good writer to a great writer by getting suggestions for revision from an AI Chatbot … (but) it is potentially helpful,” he said. The bot draws from text all over the internet, - its writing is not going to reflect high art - but it could potentially help someone reach proficiency. McKinney also is excited about the abilities of GPT-4 to serve as an assistant, especially for those who wouldn’t otherwise have access to the resources it can provide. “In an ideal world, Chat-GPT would be used to improve equity both in the classroom and as students are struggling at home,” she said. Students with more privilege have access to tutors and technologies — in McKinney’s eyes, GPT-4 could “level the playing field.”

Discussions about GPT-4 and its analogues tend to focus on possible uses for students, but the automation it provides can help teachers concentrate more on actually teaching rather than producing the sheer volume of material needed to run a successful class. McKinney uses differentiated instruction in her classroom, where she tailors the content she teaches to best resonate with her students’ interests. This strategy works best when McKinney has access to a lot of varied material, the creation of which takes a person a long time. “(GPT-4) is a great starting point for me to generate these examples very quickly,” she said. In McKinney’s ideal world, AI technologies could handle some of the grunt work, as it were, of teaching, allowing her to focus on what matters more, like planning lessons and giving feedback. “There's a lot that teachers do that we just don't have time to do all the time. And this gives us the potential to be able to do that,” McKinney said.

These technologies are, some fear, expanding faster than rules and regulations can keep up with. “I know that AI is supposedly doing really great work … (but) I would say I’m pretty alarmed by some of that I’m hearing,” said Salaverry, who is perplexed by what he sees as a lack of control within the entire tech industry. “They don’t really know what they’re unleashing into the world,” he said. McKinney concurs, thinking that what we see today is just the beginning of an artificial intelligence explosion. “I (believe) that it’s going to be unavoidable … that is concerning, to figure out how you keep your … agency while also not ignoring the tools that are available to you,” she said. These questions are demanding answers now more than ever.