ISSL Blog: ChatGPT in Schools

By Chris Cunningham, Ph.D. | Head of School, Whitfield School

Some of you have perhaps been following the news of an extraordinary new piece of technology: a so-called “generative artificial intelligence” or generative AI named ChatGPT. For many years, we’ve had software that was increasingly good at more and more complex tasks, like playing Chess or Go—or winning Jeopardy. But developers have struggled to create AIs that could generate language in ways that seemed creative and original, that didn’t simply copy or recombine pieces of text written by human beings. ChatGPT does just that—and more.

I had been reading about ChatGPT since the late fall, but I hadn’t really had time to try it out and play with it until our recent winter break. Here's a selected collection of my experiments, including work for English, History, French, Science, and Math. As you can see, ChatGPT can create email communication, arguments about literary texts and historical events, original poetry and fiction, fluent French (among other languages), cogent explanations of scientific concepts, and solutions for common problems in Mathematics, from Algebra to Calculus (it even shows its work!). In short, it is extraordinarily good at many of the things that we ask our students to do each day. [It does make mistakes, even, surprisingly, in Mathematics, and it can struggle to do tasks that require an understanding of the physical world.]

As you can imagine, a number of academic departments at our school have already been talking about both the challenges and the opportunities that AIs like ChatGPT present to teachers and to school as we know it. ChatGPT and the even smarter and more sophisticated iterations of AI that will follow in the months and years ahead challenge our work in both practical and profound ways, forcing us to confront questions both small and large:

How do I know if the work I'm grading—writing in any language and in any subject, math problems, written code, etc.—was done by my students?

How do I grade work that was entirely or partially written by an AI?

If using a spellchecker and grammar checker isn't cheating, is it cheating for a student to submit writing that an AI has edited?

When I was in high school, I was taught how to calculate the value of logarithms by hand, using tables in the back of my textbook. No one teaches that anymore because a calculator can do it more quickly and more accurately (same for square roots, trig values, etc.), and Math teachers decided long ago that it was more important for students to spend their time understanding how to use logarithms. Our current students do all kinds of things on their graphing calendars that I was tested on when I was in high school and college. If an AI can do basic (or even not basic) Algebra for us, how is that different? Do we stop teaching Algebra and focus on how to use Algebra—and what does that even mean? And so on…

As you can see, it doesn't take very long before we start getting to some fundamental questions about schools themselves, forcing us to think not only about How we teach (and assess) but also What we teach and Why.

For example, in my discipline, we've always argued that students need to learn how to write, at the very least, for the purposes of communication. In the professional world, they'll need to write reports, send emails, etc., even if they never write another essay about Hamlet in their lives. But an AI does basic communication pretty well (here is the welcome-back-to-school note that ChatGPT composed for me to send to faculty and staff). How much regular professional communication will be done by AIs in five years? Does that mean that we should stop teaching writing? If not, if we believe that writing is worth teaching—and, to be clear, I think it is—then how does ChatGPT and its successors change how we think about the teaching of this basic skill?

In his dialog Phaedrus, Plato laments the invention of written language, arguing—correctly—that people will no longer be as good at remembering things: “It will implant forgetfulness in their souls. They will cease to exercise memory because they rely on that which is written, calling things to remembrance no longer from within themselves, but by means of external marks.” Our memories—our memory skills—are paltry compared to those of people who live or lived in cultures without the written word. Similarly, Europeans in the 15th and 16th centuries were equally concerned—and for good reason!—about the socially, religiously, and politically disruptive effects of movable type and the democratization of the printed word.

We are living through a similar moment in history. In the past 50 years, robotics has profoundly disrupted industrial labor, rendering repetitive, non-creative, physical work all but obsolete. We have to assume that ChatGPT and other AIs will be similarly disruptive to many types of repetitive, non-creative intellectual labor.

To be clear, I still think that there will be things for people to do in the future. In the same way that our culture has come to re-value individually crafted objects and artisanal work—witness the emergence of makerspaces, Etsy, etc.—I suspect that we will find new and special value in intellectual handiwork as well. In the end, generative AIs will be just one more tool that we will teach our students how to use, no different than a hammer or a spellchecker.

Creativity, critical and analytical thinking, ethical judgment, teamwork and collaboration, leadership: More than ever, these will be the skills our children will need to thrive academically and professionally in the century ahead.

Parents, I encourage you to try ChatGPT for yourself and then talk to your children about it. To be sure, it will tempt some students at some schools to take shortcuts, to let someone else—in this case, something else—do their work for them. So it’s important that we have conversations with our students about the goal of their education—that is, to develop the skills and deep understandings they will need for a productive and meaningful life well-lived.

P.S. Full disclosure: I knew that Plato had argued against writing but used Google to confirm the name of the dialog and the quotation. As you can see, this example not only validates Plato's argument but also supports those who argue that the technology of the internet has made certain kinds of memory less necessary. On the other hand, it also reinforces the need for education. I needed to have enough education to have been exposed to the idea that Plato made this argument (and to know who Plato was), but didn't necessarily need to know which dialog—or to have read that dialog!—much less memorized it.