top of page

I’m not (too) afraid of AI

Before my students begin a writing assignment, I will sometimes write on the board: “Details, details. There is originality and truth only in the details.” I recorded those words in a journal years ago, followed by “Stendhal” in parentheses. I hadn’t read (and still have not read) anything else that Henri Beyle, the man known as Stendhal, wrote, but I do love that quote.

It comes to me when I’m attempting any of my own (far too infrequent) creative writing, and it came to me the other day when I was playing around with ChatGPT. I’ve read a million and one articles about how AI tools like ChatGPT and Bing will forever change education and will put us on a path to ever-more rampant cheating and plagiarism. I suppose that’s possible, though I think it’s also possible that the proportion of people - adults and students alike - who are apt to cheat or purposely plagiarize might hold steady through each technological innovation. The same proportion of students who read SparkNotes rather than 1984 will probably ask ChatGPT to summarize the story. The same proportion of Lascaux children probably copied their neighbor’s bison in cave painting class.

In any event, I signed up for ChatGPT and started to play around, wondering what kind of writing would come out in response to a variety of prompts. Because it was a sample question on the home screen, I let ChatGPT explain quantum entanglement to me in simple terms (fascinating); I asked it to explain when the subjunctive is used in French (helpful and clear); and I asked it to identify themes in The Samurai’s Garden (a novel my class is reading). Overall, the responses it generated were reasonable and readable.

What it returned to me in response to my question about themes in The Samurai’s Garden was a decent, though flawed, introductory sentence along with a numbered list of briefly described themes (all of which I could also find through Google searching). ChatGPT’s opening sentence read: "The Samurai's Garden is a novel by Gail Tsukiyama, set in Japan and China during the 1930s and 1940s, and follows the story of a young Chinese man named Stephen who is sent to a seaside village in Japan to recover from tuberculosis.” I don’t love the syntax of that sentence, and the story takes place entirely from 1937-1938, but it does model the need to situate the reader before diving right in. What’s the work, who’s the author, what is the work generally about? The themes it listed were fine but lacking in detail and specific supporting passages. In short, it didn’t write an acceptable essay response, but if I were a student floundering with a blank page, it might have given me a jumpstart.

Next in my ChatGPT exploration, I chose a few sentences that my students had written when responding to Billy Collins’ poem, The History Teacher, and asked ChatGPT to analyze them. Here’s one, as written by the student: "In this section it is made abundantly clear that despite the history teacher’s deception, the kids on the playground are still a bunch of jerks but now armed with a false narrative."

ChatGPT responded that the sentence might be better written: "In this section, it is made abundantly clear that despite the history teacher’s deception, the kids on the playground are still behaving badly, but now they are armed with a false narrative."

I don’t think it would be cheating for a student to get that kind of input from AI. Contrasting the two sentences could lead to some interesting discussion ideas. Why change “still a bunch of jerks” to “still behaving badly”? Is it always necessary to change informal language to more formal language? Does adding a subject in the final clause make it more readable? I threw in a few more sentences and a few paragraphs for analysis. In many cases, ChatGPT successfully corrected grammar and even found ways to correct run-on sentences. (ChatGPT could not, however, identify when a missing word caused an opening clause to make no sense; it simply added the missing comma.)

Because it was report card time when I was playing around, my final ChatGPT experiment was to ask it to write a complimentary report card comment for a ninth grade literature student. It produced this:

"[Student Name] is a highly engaged and thoughtful reader who consistently demonstrates a strong understanding of literary texts. Their contributions to class discussions are insightful and demonstrate a deep understanding of the material. They consistently produce well-written and well-organized written responses that reflect their close reading and critical thinking skills. Keep up the great work!"

Again, not terrible, and better than the drop-down comment my son once brought home from middle school: Your son/daughter is a pleasure to have in class. Keep up the good work!

The ChatGPT-generated comment was more informative than the one about my son/daughter, but Stendhal would have identified the weakness: There is originality and truth only in the details. Neither of these comments contain anything that communicates a genuine connection or relationship, and neither contains the specificity that would make the comment unique to the student receiving it. But, as a teacher, could I use that framework as a comment-writing jumpstart? Could I add the detail and specificity that would make it more meaningful? And if I were writing eighty or a hundred comments, could ChatGPT help me spend more of my time on those more meaningful bits (details, details), the parts that make the student feel seen and deepen our connections?

Chat GPT is a rabbit hole and I spent quite a bit more time asking more and more detailed and specific questions. For certain question types, I did find ways determined students could cheat, with AI producing a few passages I might believe were theirs. Those passages also passed the plagiarism checker I use. But plagiarism checkers aren’t foolproof; they’ve never, for example, been able to nab the parent who writes a student’s essay. In addition, the AI passages contained no in-text citations or direct quotations from the text. For a student who hadn’t read the work, it would be time-consuming to add those elements to the AI paragraph.

It’s clear that AI is going to open another realm for schools to think about. Some days I think we might be happier and healthier with card catalogs, electric typewriters, and no Instagram photos of each other’s dinners (not to mention the darker elements of social media). Still, along with my worries about the perils of AI in the classroom, I do see intriguing possibilities for both students and teachers.

What do you think? We’d love to hear how your school is responding.


bottom of page