The past month or so has reminded me in some ways of the early days of the internet when people would send the latest thing by email with breathless notes: “Can you believe we can send photographs to each other?” I remember how enthusiastically my friends and family used to forward me things all the time, but then the novelty of doing that wore off, and the volume of forwarded emails slowly declined. The arrival on the scene of ChatGPT
, a chatbot launched back in November by OpenAI, changed all of that. Suddenly my inbox and texts were flooded again with articles, news items, podcasts, messages, and other information about ChatGPT:
“It’s going to change the world!”
“This is the end of high-school English and History as we know them.”
“What are you going to do about this?”
“Schools need to ban this right away!”
In case you’ve missed it, ChatGPT is an artificial intelligence-based tool that will write very natural language in response to a prompt. It works using what’s known as a “Large Language Model,” in which the system is trained through machine learning on an extremely large database of existing written language. With the help of input parameters, the chatbot can predict what words will naturally come next and put together a coherent and original piece of writing that won’t appear to be plagiarized or show up anywhere on the internet.
The debates over ChatGPT and its role in the classroom are rife with doom and gloom and a fair amount of prognostication. Some are panicked about its impact and take a reactionary stance: for example, New York City recently announced that it was banning ChatGPT
in all of its public schools. To me, the debate is reminiscent of discussions early in my career about the impact of graphing calculators on math teaching, where many pundits worried that students would use the calculators to cheat. Schools and teachers who embraced the technology, however, eventually came to realize that the graphing calculator allowed students access to a higher level of thinking. I remember when I started using the graphing calculator to introduce optimization, a topic that used to wait until calculus, with my Algebra II students.
Of course, many would argue that graphing mathematical functions is a lower-level skill than writing, which most of us spend all of our lives learning to master. But consider the provocative arguments of Conrad Wolfram in his book The Maths Fix: An Education Blueprint for the AI Age
. Wolfram, one of the co-founders and currently the CEO of Wolfram Research, argues that we should relegate the entire academic discipline we now know as “math” to the place currently occupied in most schools by subjects like Ancient Greek. Wolfram asserts that we should replace math with an entirely new discipline that focuses on computational thinking. Why not accept that computers can do math calculations faster and more accurately than humans? And, if we accept that fact, why are we not teaching students to do the higher-level thinking that is necessary to program the computers to do those calculations for us?
Although I am very sympathetic to Wolfram’s arguments, I also recognize that there are quite a few practicalities to consider before we take such a radical step in our schools. Similarly, I’m not arguing that we should simply open the doors to ChatGPT and let students use it for anything and everything. If we are truly interested in assessing what students can write without the assistance of technology, we should be able to control the in-class conditions under which students do their writing. At other times, however, we should take advantage of opportunities to push our students to engage with and think critically about ChatGPT and other similar tools. Since its launch, critics have discovered that, while it can write well, ChatGPT can also be spectacularly wrong. For example, one writer
gave the system a prompt from his daughter’s history class about whether Thomas Hobbes believed in the separation of powers. Although the philosopher did not believe in or advocate for a governmental system of checks and balances, ChatGPT wrote convincingly about how he did. It turns out that because Hobbes is almost always written about alongside John Locke, who did believe in the separation of powers, the AI incorrectly associated Locke’s ideas with Hobbes.
I believe that schools and teachers should challenge students to think critically about the platform. In a history class, we could challenge students to “trick” the chatbot into writing a factually inaccurate essay and then have the students critique the bot’s work, grounding their analysis in the historical contexts they have learned about in class. In English, we could ask our students to use the tool to write a first draft and then focus more intensively on the revision process that ensues. One teacher from Oregon named Cherie Shields
has quickly refocused her classroom around ChatGPT in innovative ways. For example, although more advanced tools are likely to come along soon, ChatGPT doesn’t write particularly lengthy or nuanced essays without a lot of careful and iteratively-applied inputs. Shields pushes her students to develop questioning strategies that can coax a more sophisticated essay from the tool. Shields has also experimented with having her students use the tool to write outlines and even to give the students near-real-time feedback on their writing. None of this replaces the work and thinking Shields’ students have to do in class or even on assignments, but her students are learning to use this powerful tool appropriately.
At Field, we’ve never shied away from new ideas or from thinking critically about how students can and should be learning. ChatGPT is the latest challenge to this thinking. I, for one, am thrilled and invigorated to be a part of a school community where we embrace and discuss ideas like this as a faculty and with our students, and where we view something like ChatGPT not as a threat but as an opportunity.
Page Stites is the Upper School Director at Field and did not use ChatGPT to help him write this blog post!