Exquisite AI Corpse

Experiment in the face of the unknown.

Assignment Details


AI Theme




Learning Objectives

Collaborate with a variety of LLM tools to experiment with language and reflect on AI’s “collective unconscious”


In the 1920s, Surrealist artists and writers invented a creative experiment inspired by old parlour games called “Exquisite Corpse” in which participants would collaboratively draw the head, torso, and lower limbs of a body or creature on a piece of paper. The key gimmick of this game was that the paper would be folded such that no one could see the preceding drawings. Taken together, these unique parts constituted a surprising new image that could not be achieved individually.

  1. For this assignment, students will collaborate with at least three AI text generators. For instance, they might use ChatGPT, Google Gemini, and Microsoft Copilot. Depending on the course goals and the context in which this assignment is given, students can either work on a piece of creative writing or an analytical essay. They might produce a short story, poem, or response paper related to a specific course topic. 
  2. If students are working on short stories, for instance, have them start by writing the first paragraph or page of their story independently. Then, students will turn to one of the AI text tools and prompt it to continue the next portion of the story. A prompt might entail something along the lines of the following: “We’re going to play a game of exquisite corpse. I wrote the first part of a short story and I would like you to write the next part. Can you please contribute 250 words and continue the story based on the following sentence: As he looked out the window waiting for her to return, the sun started crying and a dense flock of starlings flying in from the east alighted on the field.”
  3. Students will then continue the story by writing the next portion themselves. This works best if they can avoid reading the entirety of what the chatbot wrote! Encourage them to cover the screen or scroll to the bottom of the page to read just the last sentence.
  4. Turning to the second AI chatbot, students will then input the same prompt and the last sentence of the preceding portion they wrote, and so forth until three different LLMs have contributed a portion to the story, interspersed between the student’s own writing. Students should have contributed approximately 3 pages (roughly 1,000 words) of their own writing to the final result.
  5. Once the “exquisite corpse” has been completed, students will write a brief 1-2 page reflection on the experience. Here are some questions they might consider:
    • Do you think the final product is surprisingly coherent, beautifully fragmented, or disappointingly dull?
    • How did it feel to produce something creatively (or analytically) in the face of the unknown?
    • Does it appear to be the case that one chatbot did a better job of making a more interesting contribution to the story/poem/essay than the others? If so, which one and why?
    • What do you think this experiment reveals about the “collective unconscious” of AI, or the ways in which a variety of LLMs may exhibit similar or divergent stories or sentiments?
    • What was the experience like for you giving over your story or essay to the LLMs? Did it feel like collaborating with another human, or entirely distinct?
    • Did anything about the experience surprise you, either the contributions by the LLM or your own reactions/contributions in turn?
    • Might you use a technique like this in the future? What risks, if any, do you imagine there would be with this kind of co-writing?
    • Who is the author of your piece? 


DuPont Analysis of Coca-Cola

Conduct a DuPont analysis of Coca-Cola using real-world data, comparing and contrasting your results with an analysis conducted by Microsoft’s chatbot Copilot.


How to Train Your AI Dragon

Empower educators to craft innovative activities by training AI tools to enhance student learning outcomes, generate educational content, and design evaluation methods.