Composing with AI

Paper Assignment: What AI Left Out

In the early days after the release of ChatGPT, November 30, 2022, a day that shall live in infamy, I was trying to find a way to revamp my assignments so they could not be easily completed by an LLM. Like most of my writing professor colleagues, our winter breaks had been full of yet-unseen levels of interest in the teaching of writing, specifically around the question: What are you going to do about ChatGPT? It seems that the world had all discovered LLMs at once, even if they had been around for years before.

While my colleagues suggested that their assignments were AI-proof, I was not so convinced. What if we let the students use LLMs or -- perhaps more provocatively -- invited them to use it? Their response: a volley of boos and tomato emojis. Reconsidering, I asked, what if students wrote responses to what the AI generated?

So, I issued a paper assignment that I have since seen many peers using in which students begin by asking the LLM to produce a paper based on the assignment prompt for the class. Then, the paper they write is an evaluation of that paper, primarily based on what has been omitted from the paper. Conceivably systems such as Claude and ChatPDF, that focus more on analyzing input texts, could write the response paper, so I leaned heavily into the students’ lived experiences, their particular insights that an AI would not have access to. The assignment is a way to celebrate that unique offering each student brings to a topic.

While the first round of papers were focused a bit too much on critiquing the writing qua writing, a good portion of the students offered content drawn from their perspectives and experiences. Revising this assignment for future classes, I would make sure to emphasize the writing about omitted content rather than critiquing the writing.