Esther Morgan-Ellis is Associate Professor of Music History and Assistant Director of Academic Engagement.

If you are not familiar with the capabilities of ChatGPT, start by visiting the website now and trying it out. You’ll have to take a few minutes to set up a free account. After that, try entering some questions that you might ask your students. They can be essay prompts, discussion starters, factual queries, analytical tasks—anything. Ask it to summarize a text or compare and contrast scholarly views. Tell it how many words you want, require citations, and suggest a style. You might be surprised by what ChatGPT can do.

Esther Morgan-Ellis

ChatGPT, which was unveiled late in 2022, does not work by searching the Internet. Instead, it has been trained on an enormous text dataset to essentially predict what word should come next in a sentence. The text it produces is “original,” in the sense that it is newly constructed and cannot be traced to a specific source.

Currently, ChatGPT has some striking limitations. It does not fact-check, and it can provide startlingly inaccurate information. It often “invents” citations. It is unoriginal and mainstream in its perspectives, since it works by stringing words together in the most common sequence. When asked to produce a lengthy text, it often becomes repetitive. Likewise, two answers to the same question, while unique, are likely to share many similarities, including identical sentences. The default style is bland and easily recognizable to an experienced reader.

However, it is a tool under rapid development, and these limitations are already evaporating. We can assume that AI programs like this one will soon be able to complete almost any writing task at an expert level.

ChatGPT has obvious implications for education at all levels, and an enormous amount has already been written on the subject. The most common question educators are asking is, “How can I prevent my students from using ChatGPT to cheat?” Responses have ranged from doomsaying to enthusiastic embrace of the new technology. I have my own perspective, but I have also read the commentary and participated in extensive in-person and online discussions with other teachers in my field. I find that there are essentially four approaches to confronting ChatGPT with your students:

  1. Prohibit the use of ChatGPT

This is the simplest response in a way, but it places a significant burden on you to identify and respond to infractions. First, it is difficult to “prove” that a student has used AI-writing technology. A detection program followed closely on the heels of ChatGPT, and it can help you to identify AI-generated text: https://openai-openai-detector.hf.space/. However, this approach will not work indefinitely.

There are also ways to identify AI use manually. You might require students to complete a short in-class writing assignment at the beginning of each semester. This then becomes a sample against which you can compare later submissions. Most students have a recognizable style, complete with grammatical errors, and it is obvious to the professor when they deviate from it. This approach might not work in a large or online class, but I have no trouble recognizing my students’ writing in the context of a small in-person class with weekly online writing assignments.

For writing assignments on which I prohibit the use of ChatGPT, I have a simple policy: If I suspect that the student has used AI assistance, I reserve the right to schedule an oral examination in which I will query their knowledge on the topic and ask them about their writing process. I have not yet had to follow through on this threat, and I don’t expect that I will have to. My students were horrified at the idea.

  1. Disincentivize the use of ChatGPT

This is my preferred response. Most students are well-intentioned, and they cheat because they don’t feel that they have a viable alternative. In my experience, academic dishonesty is almost always associated with high-stakes assessment and the absence of scaffolding. Students cheat because they are faced with irreconcilable failure and/or have run out of time to do the work.

I disincentivize cheating by making all assignments revisable. This includes both large-scale writing assignments, which have components due periodically throughout the semester, and weekly short-answer responses to questions, which can be revised until the student earns full credit. ChatGPT produces pretty good work in response to most of my assignments. However, it always needs to be revised. A student will have an easier time revising something they actually wrote—something that reflects their understanding and perspective.

I also talk to my students at great length about how important writing is both as a skill and as a process for thinking and learning. I believe profoundly in the value of writing. This is not a game where my students try to outsmart me and I try to catch them. We are on the same team. I want them to learn through writing and to become better writers, and I clearly and frequently explain my reasons for asking them to write (an approach often styled as TILT, or Transparency in Teaching and Learning). This, in combination with the revision process, has so far eliminated any concerns regarding ChatGPT in my classroom.

  1. Negate the value of ChatGPT

ChatGPT works very well for many types of writing assignments, but not all. Currently, I have students working on a project for which they must write historical narratives based on evidence from old issues of the Dahlonega Nugget. That is not something that ChatGPT can do, or even assist with. It has no access to the sources they are using and no “knowledge” of the marginal actors or events they are tasked with describing.

The challenge of setting writing tasks that are beyond the ability of ChatGPT is unique within each discipline, but there is always a way to craft an assignment to be so individual, so highly specialized, or so specific to your classroom context that AI will be of no use. These assignments can also be the most meaningful to students. It is exciting for them to write something that only they can write.

  1. Embrace ChatGPT as a learning and writing tool

While I have not tried this approach myself, many of my colleagues have put forth ideas for building assignments around ChatGPT. These approaches require students to use the tool. Students then reflect upon or manipulate the output. Here is a list of possibilities I have come across:

  • Have students critique or “grade” ChatGPT responses in light of what they have learned in class
  • Have students edit ChatGPT responses by making the writing more specific and adding citations
  • Use ChatGPT to produce first drafts of essays, which students then build upon
  • Have students ask ChatGPT a subjective question, and then argue against whatever side it takes
  • Have students revise ChatGPT prompts until the answer returned includes exactly the desired information, which is an exercise in critical thinking
  • Use ChatGPT to teach students about authorial voice, since it can replicate a wide range of writing styles
  • Teach students to properly cite ChatGPT—just ask, and it will tell you how!

By requiring students to use ChatGPT and drawing attention to both its limitations and its potential as a tool, you teach students that they are better than AI technology. They have something to bring to the table above and beyond what machine learning can do.