Live chat: A new writing course for the age of artificial intelligence

How is academia dealing with the influence of AI on student writing? Just ask ChatGPT, and it’ll deliver a list of 10 ways in which the rapidly expanding technology is creating both opportunities and challenges for faculty everywhere.

On the one hand, for example, while there are ethical concerns about AI compromising students’ academic integrity, there is also growing awareness of the ways in which AI tools might actually support students in their research and writing.

Students in “Writing Essays with AI,” a new English seminar taught by Yale’s Ben Glaser, are exploring the many ways in which the expanding number of AI tools are influencing written expression, and how they might help or harm their own development as writers.

“We talk about how large language models are already and will continue to be quite transformative,” Glaser said, “not just of college writing but of communication in general.”

An associate professor of English in Yale’s Faculty of Arts and Sciences, Glaser sat down with Yale News to talk about the need for AI literacy, ChatGPT’s love of lists, and how the generative chatbot helped him write the course syllabus.

Is the object of your new course to explore the way ChatGPT can influence written expression, or is it to instruct students in how to use it ethically? Or both?

Ben Glaser: It’s more the former. None of the final written work for the class is written with ChatGPT or any other large language model or chatbot, although we talk about using AI research tools like Elicit and other things in the research process. Some of the small assignments directly call for students to engage with ChatGPT, get outputs, and then reflect on it. And in that process, they learn how to correctly cite ChatGPT.

The Poorvu Center for Teaching and Learning has a pretty useful page with AI guidelines. As part of this class, we read that website and talked about whether those guidelines seem to match students’ own experience of usage and what their friends are doing.

Do you think students are confused about how to use ChatGPT ethically? And do you get the sense that they’re using it quite often?

Glaser: I don’t get the sense that they are confused about it in my class because we talk about it all the time. These are students who simultaneously want to understand the technology better, maybe go into that field, and they also want to learn how to write. They don’t think they’re going to learn how to write by using those AI tools better. But they want to think about it.

That’s a very optimistic take, but I think that Yale makes that possible through the resources it has for writing help, and students are often directed to those resources. If you’re in a class where the writing has many stages — drafting, revision — it’s hard to imagine where ChatGPT is going to give you anything good, partly because you’re going to have to revise it so much.

That said, it’s a totally different world if you’re in high school or a large university without those resources. And then of course there are situations that have always led to plagiarism, where you’re strung out at the last minute and you copy something from Google.

What would you say is the proper role for AI in writing?

Glaser: First of all, it’s a really interesting thing to study. That’s not what you’re asking — you’re asking what it can do or where does it belong in a writing process. But when you talk to a chatbot, you get this fuzzy, weird image of culture back. You might get counterpoints to your ideas, and then you need to evaluate whether those counterpoints or supporting evidence for your ideas are actually good ones. There’s no understanding behind the model. It’s based on statistical probabilities — it’s guessing which word comes next. It sometimes does so in a way that speeds things along.

If you say, give me some points and counterpoints in, say, AI use in second-language learning, it might spit out 10 good things and 10 bad things. It loves to give lists. And there’s a kind of literacy to reading those outputs. Students in this class are gaining some of that literacy.

Are you saying it can be used as more of a tool for learning before writing?

Glaser: I don’t love the word brainstorming, but I think there is a moment where you have a blank page, and you think you have a topic, and the process of refining that involves research. ChatGPT’s not the most wonderful research tool, but it sure is an easy one.

I asked it to write the syllabus for this course initially. What it did was it helped me locate some researchers that I didn’t know, it gave me some ideas for units. And then I had to write the whole thing over again, of course. But that was somewhat helpful.

So it can be a jumping off point?

Glaser: It can be. I think that’s a limited and effective use of it in many contexts.

One of my favorite class days was when we went to the library and had a library session. It’s an insanely amazing resource at Yale. Students have personal librarians, if they want them. Also, Yale pays for these massive databases that are curating stuff for the students. The students quickly saw that these resources are probably going to make things go smoother long-term if they know how to use them.

So it's not a simple “AI tool bad, Yale resource good.” You might start with the quickly accessible AI tool, and then go to a librarian, and say, like, here’s a different version of this. And then you’re inside the research process.

Are there creative ways to use ChatGPT in writing?

Glaser: One thing that some writers have done is, if you interact with it long enough, and give it new prompts and develop its outputs, you can get something pretty cool. At that point you’ve done just as much work, and you’ve done a different kind of creative or intellectual project. And I’m all for that. If everything’s cited, and you develop a creative work through some elaborate back-and-forth or programming effort including these tools, you’re just doing something wild and interesting.

Anything else you’d like to say about the course?

Glaser: I’m glad that I could offer a class that students who are coming from computer science and STEM disciplines, but also want to learn how to write, could be excited about. AI-generated language, that’s the new medium of language. The Web is full of it. Part of making students critical consumers and readers is learning to think about AI language as not totally separate from human language, but as this medium, this soup if you want, that we’re floating around in.