Author: Anton Borst, Senior Learning Designer, Office of the Provost
When ChatGPT does the work of writing for students, it derails the process of learning how to write. But what happens when AI is brought into the writing process instead of delivering the assigned product? In our first Fall 2023 TeachTalk, Incorporating Generative AI into Writing Assignments, Karen Lepri and Alexander Landfair from NYU’s Expository Writing Program shared examples of using ChatGPT to illuminate the process of writing in the classroom. The exercises they described shift the context of ChatGPT’s use from last-minute paper generator to something like an in-class thought-partner.
Professor Lepri described a film synopsis assignment that she adapted after realizing that the products that the assignment asked for could be produced easily by ChatGPT. In class, students wrote three- and one-sentence summaries of their chosen film and then asked ChatGPT to do the same. Then they compared the outputs, noting, for example, how their own summaries might better serve–in tone, style, and content–the one-page synopsis they now planned to write. But the large-language-model-produced summaries also gave students clues as to what the larger cultural conversation around their chosen film looked like. On their own, a student might not know to refer to their film as “absurdist” (for example), but seeing the term produced by a probability-based language generator would prompt them to look it up. The comparison activity gave students more authority over their topic and their writing, which is one of Lepri’s primary learning goals for the course.
Professor Landfair, while underscoring the broad possibilities for misuse of generative AI in college coursework, explained how he, too, used ChatGPT to help students meet one of the primary learning goals for his writing course: developing a richer understanding of the writing process. His exercise requires students to train ChatGPT to follow a set of rules–in this case “Roger’s rules” of rhetoric. Students had to write and rewrite their explanation of the rules to make them followable by ChatGPT, an activity that highlighted the complex nature of writing, the need to understand and address a particular reader, and the acts of reframing and rearranging often involved in effective written expression.
During the conversation, Lepri and Landfair agreed on ensuring that students, when interacting with generative AI during the writing process, assume a position of authority over the tool. I came to a similar conclusion this summer when teaching academic writing at the School of Professional Studies. Several times during the course, I asked students to use and assess ChatGPT’s capability for a specific task. For example, after reading and discussing an article in class, students, in small groups, evaluated several summaries of the same article written by ChatGPT. In the process, they identified the limits of ChatGPT while recognizing the nuance possible in summary, noting, among other things, how ChatGPT’s selection of detail and phrasing represented the article in ways they did not alway agree with. This and other conversations situated generative AI as a limited tool for specific tasks, rather than a one-stop source.
You can watch this TeachTalk in its entirety, review the slide deck, and register for future sessions on generative AI on the TeachTalks page. NYU’s Office of the Provost will continue providing support for navigating AI and teaching and learning throughout the year. For the latest programs, guidance, and resources, visit:
Join us for the NYU Teaching & Learning Generative AI Virtual Conference & Pre-Conference Workshops on October 26–27, 2023.
And, if you’d like to continue the conversation weekly, log on to NYU’s very active Google space devoted to teaching and learning with AI tools. Email email@example.com to join.
Resources shared during the TeachTalk:
Classroom Policies for Generative AI Tools – a crowdsourced document