Students Critique a ChatGPT Essay
A Classroom Experiment
Historians today find ourselves on the bleeding edge of the artificial intelligence (AI) revolution. It feels like generative AI is everywhere these days, even in places where it’s not welcome. Chatbots like Open AI’s ChatGPT have made disruptive appearances in law, real estate, Hollywood, and of course the classroom, offering up simultaneous promise and peril. While Bill Gates described AI as the “most important tech advance in decades,” many tech visionaries (including Gates), AI developers, and even the White House have called urgently for guardrails on AI’s breakneck development.
The meteoric rise of chatbots raises questions that, if left unanswered, put history educators in a bind. How can we meaningfully teach historical skills like writing and argumentation if our students can have ChatGPT draft an essay for them in seconds? And how can we assess our students’ understanding of course content if ChatGPT can conjure up instant answers to essay prompts with no learning required? Unsurprisingly, universities, departments, and individual instructors are in a mad scramble to formulate AI policies addressing these urgent questions.
When I first encountered ChatGPT shortly after its November 2022 launch, I admittedly felt a knee-jerk impulse to ban it and forget about it. Like many of my colleagues, I was deeply worried about its potential for academic dishonesty (and I still am). But it was obvious that AI is here to stay. More importantly, as I practiced using ChatGPT, I realized that effective AI use is a learned skill, one today’s students desperately need to learn as chatbots proliferate. If we envision history courses as opportunities to teach the historian’s skill set, we would do well to incorporate new technology like AI into our courses. So, I decided to figure out how I could effectively use ChatGPT in the classroom for my students’ benefit.
Conversations with my students about the chatbot revealed their curiosity about both its uses and limitations. So, I created a classroom experiment to test ChatGPT’s ability to generate authentic, accurate historical essays. I chose to do this experiment live and in person in a spring 2023 upper-level undergraduate history course, but the exercise could also work as a take-home assignment. This particular class, Frederick Douglass’s America, familiarizes students with the life and times of the famous 19th-century abolitionist and introduces the historical research skills that students will need to complete an undergraduate capstone. As we were approaching the semester’s midpoint, I also designed the exercise as an opportunity for students to reflect on what they had learned about Douglass’s life thus far.
Effective AI use is a learned skill, one that today’s students desperately need.
To that end, I asked ChatGPT to “write me an essay about Frederick Douglass and the Civil War.” The program generated a relevant, if basic, biographical sketch of Douglass without “hallucinating” any fake information, a notorious shortcoming of ChatGPT. The essay lacked an argument or references, but I didn’t ask for either, so no harm done. Then I graded the essay, which earned a D by my rubric.
Next, I shared the essay with students via Google Docs. (It can be read in full here, along with the lesson plan.) In groups, I asked them to read the essay and identify any sections that needed editing—additions, deletions, or revisions. For 20 minutes, students tackled this assignment in groups and focused on a theme of their choice: Douglass’s life under slavery, his wartime experiences, or his Reconstruction-era activism. Students added their responses in the document using comment bubbles.
Finally, we debriefed. I asked students to describe a factual error they identified and an edit they made to correct or contextualize the essay. Then I asked students to reflect on their biggest takeaway.
This exercise was a smash hit, with terrific student engagement and clear payoffs. Students jumped at the chance to showcase the knowledge they had learned in class to fact-check the essay. Ultimately, they made a few deletions, some additions, and, most importantly, used their higher-order thinking skills to evaluate the essay and add much-needed context to its key points. For example, where ChatGPT matter-of-factly stated that Douglass was born in 1818, students observed that, like many enslaved African Americans, Douglass spent most of his life without knowing his birthdate. This and other legacies of slavery are all too often overlooked by chatbots, which lack contextualization and fine-tuned historical research skills.
Students jumped at the chance to showcase the knowledge they had learned in class to fact-check the essay.
Instead of merely regurgitating information on a midterm exam, students enjoyed probing the limits of what ChatGPT can realistically do—and not do—hopefully disincentivizing future academic misconduct. In their reflections during the debrief, students described how ChatGPT generates essays that seem human-produced and authoritative, but the significant factual errors and lack of analysis in its text often make a poor substitute for how students themselves have learned to write. They also sharpened their writing and editing skills in an engaging activity that felt more pragmatic than a midterm exam.
I learned a lot from this exercise too, which can be adapted readily for a variety of course formats. I came away recognizing that if I had given into the initial impulse to ban ChatGPT, my students would have been robbed of a crucial learning experience to prepare for a world that is rapidly embracing AI. Instead, by incorporating the chatbot into the classroom, my students gained a better understanding of what AI can (and can’t) do, while also sharpening their historical skills and applying their content knowledge.
After all, AI is not disappearing anytime soon, and our students are already using it. Many historians likewise have embraced digital literacy as one of the most essential skills we can teach students, but if educators bury our heads in the sand and attempt to ban AI from the classroom, crucial learning opportunities are lost. So why not bring AI into the classroom and harness its potential to reinforce historical skills and content?
Jonathan S. Jones is an assistant professor at James Madison University. Find him on X (formerly Twitter) @_jonathansjones.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. Attribution must provide author name, article title, Perspectives on History, date of publication, and a link to this page. This license applies only to the article, not to text or images used here by permission.
Please read our commenting and letters policy before submitting.