Stephen’s Thoughts On… Artificial Intelligence
EDITOR’S NOTE: Over the campaign I write about topics important to Atlanta Public Schools. The third of this “Stephen’s Thoughts” series is a note about the use of Large Language Models like ChatGPT in schools.
When I was in grad school my wife supported us financially through her job as a server so I could be a student full-time. Over breaks I also worked for the same restaurant as an “expo” or someone who transported the food from the kitchen to the tables. This position required keeping full all the elements of the meal that were not made by the cooks. Before the shift began expos were supposed to check that we had enough chips, toppings, sauces, etc. One time there wasn’t enough cocktail sauce so I asked where to find it. I was pointed to one of the managers who told me that we make it in-house.
This manager, who went by “Chef,” walked me through the ingredients and amounts in order to make the cocktail sauce. I thought it was strange that he was telling me all this because who cares I just needed the container to be filled I didn’t need a lesson. Honestly I wasn’t planning on following him at all. I was headed to the walk-in and he happened to be headed that way too and he started talking about how to make the sauce. Maybe it was his recipe and he was proud? I thought it would be rude to ignore him so I followed him until it was finished.
A few shifts later we ran out but it happened during the dinner rush. I hurried over to Chef and told him and he yelled that I should go make it then since he taught me how. I didn’t have the courage to tell him that while he thought he was teaching me the recipe my brain was playing a loop of Katy Perry’s “Roar” (it had come out earlier that month).
I think about that night whenever I hear about the use of AI (Artificial Intelligence) in education.
I don’t want to keep you guessing on my overall perspective: I’m of the mind that AI as it currently stands poises a danger to students, teachers and the environment and school boards can provide thoughtful leadership on how schools can navigate the new technology.
Students can (and do) easily put the prompt of an essay into a large language model like ChatGPT and what comes out is often a passible response. It would seem like a slam dunk—the teacher asks a question and the student gives a response. Who cares how the student arrived at the answer as long as it’s right? The problem is: those students don’t know how to make the cocktail sauce. Teachers don’t require students to read Animal Farm so that they get an essay to grade. You’re supposed to read Animal Farm so that you learn about the use of allegory, the Russian Revolution, and the danger of power corrupting even your co-conspirators. The essay is just evidence that you understood what you read.
We learn things so that it changes us. Turns us into better citizens, friends, coworkers, etc. Shortcuts like AI rob our students of the necessary brain muscle-building that transform us into mature humans.
But Stephen, what if you think of AI like a search engine?
I’ve heard this argument before, that ignoring AI is like the driver who refuses GPS at his own peril. While Google Maps can get things wrong, however, I’ve yet to see an instance where it straight-up lies. AI models make up stuff (called hallucinating) all the time. They are not a reference material, and will produce plausible information that has no basis in reality because the goal of models like ChatGPT is not accuracy, but to generate “natural language” responses. It’s the difference in sounding convincing and actually being right.
I’ve personally had my child’s teacher tell me that if I wanted to help my son learn certain topics that I should type the state standard into ChatGPT and ask it to create a worksheet. Now it’s one thing if the teacher is asking AI for help and then checking to make sure the output is correct, but asking a parent to use a model that hallucinates up to 79 percent of the time to create homework is—I’ll just say it—educational malpractice.
If this technology was only being used to short-cut actual learning and tricking educators into accidentally providing erroneous work, it would still be dangerous enough to raise alarms. But my concerns don’t stop there. AI models are making it easier for students to cheat, introducing troubling privacy violations, can offer answers that cement racist or sexist bias, all the while requiring huge amounts of electricity and water at a time when our planet’s climate is experiencing significant deadly changes.
While I hold strong views, I want to admit that I’m not an expert in technology. Perhaps AI can be used in a way that helps students and teachers and doesn’t burn the Amazon forest so that kids can summarize The Great Gatsby. But I’m convinced that we are not there yet, and the students and educators of Atlanta Public Schools should not be guinea pigs in this experiment.
As a school board member, I would work with the board to first create a process to review AI practices by our learners and teachers. We need to know how widespread the issue is and by whom. While the district has a solid, if unfinished, position statement on the technology, it appears to defer to AI as some inevitable force instead of a poorly understood invention still in its infancy. If teachers are regularly counting on ChatGPT to make learning materials then our district needs to have an emergency meeting to set expectations for this technology.
We need to be leading the nation on issues like the dangers of minors using models like ChatGPT for therapy, or how to incorporate more curriculum specialists to ensure our teachers revisit the science of learning. Let’s negotiate with this new world from a position of strength—that technology must prove itself before being incorporated into our academic practices—instead of reacting to every novel product as if they have student well-being at heart.
Call me crazy but I want our teachers to teach, not a program that doesn’t know how many “r’s” are in the word strawberry.
Best,
Stephen