Made Not Only By Me: Coauthoring a Children's Book with Text and Image Generation

Brandee Easter
York University

This assignment asks undergraduate students to explore image and text generation technologies to create a short, illustrated children's book. Although text and image generation technologies are different, experimenting with them in parallel challenges students to reflect critically on the co-constitutive relationship between writing and technology. This approach was particularly useful for making "visible" how technologies can participate in, create, and sustain biases.


Learning Goals: 

  • Use experimentation with AI as a method for engaging with questions about writing, authorship, and technology
  • Interrogate biases, risks, and consequences of text and image generation
  • Reflect on the possibilities, limits, and ethical concerns with AI writing technologies

Original Assignment Context: End of large year-long, first-year Professional Writing course

Materials Needed: Accessible image and text generators for students to use

Time Frame: ~6 weeks


Introduction

This assignment was inspired by the creation of (and backlash to) Alice and Sparkle: a 12-page children’s book written by Ammaar Reshi, a design manager, in one weekend through a combination of ChatGPT and Midjourney. Like much other panicked, often rage-filled, social media discourse about these technologies, responses to Alice and Sparkle debated the boundaries, mediation, and status of writing, authorship, and originality—concepts central to “Introduction to Professional Writing,” a year-long, large-lecture, first-year course. 

In a 6-week unit on digital writing in Spring 2023, approximately 120 students produced short, scaffolded experiments in text and image generation building toward a 12-page children’s book, like Alice and Sparkle. At each stage of development—from idea generation, drafting, editing, and illustration—students work with (and against) AI tools, reflecting on what this experience reveals about the opportunities, consequences, and ethical challenges of text generation. 

Conceptually, our conversation builds from Kathleen Blake Yancey’s 2004 CCCC Chair’s Address, “Made Not Only in Words,” to place text generation technologies, and the panic surrounding them, within a larger scope of the development of writing technologies. We then use this as a frame to think about the role that technologies have always played in shaping and being shaped by literacy practices in order to move away from fear and into critical engagement. Students conclude their experiments with a reflective essay that offers suggestions for how students, teachers, and writers can or should use text or image generation in an academic setting.

One success of this assignment was that experimenting with images in addition to text prompted students to ask different questions about writing. In particular, image generation provoked more thoughtful discussions about biases, representation, and consequences than when working only with text generation. Experiments with text generation tended to raise questions about intentionally malicious actors and explicit bias, racism, sexism, and hate speech. However, the addition of image generation to our experiments seemed to offer a way to “see” how biases matter and deserve our attention. This brought discussion to questions of coded and implicit bias and followed up nicely on readings and lectures by scholars like Safiya Umoja Noble and Cathy O’Neil.

Before teaching this assignment for a second time, my greatest focus will be changes that help students not only critique but also explore the creative possibilities of text and image generation. This includes revisiting the choice of a children’s book as the final project, which was intended to provide a short and familiar genre to students—allowing them to focus on experimenting with the technologies instead of the final product—, but may have instead discouraged creativity because the genre was so short and followed predictable beats. I would also reconsider the specific text and image generators. This year’s iteration planned to use ChatGPT and DALL-E 2, but because of access issues, our tools were always in flux. Additionally, the user-friendliness of ChatGPT’s interface also worked against experimentation, leaving students with their only point of influence as prompt writing and rewriting, instead of also allowing for the selection of data sets, length parameters, or randomness, for example. Because of this, the next iteration of this course will look to generators, perhaps intentionally older and less sophisticated, that provide more opportunities for negotiating with machinic coauthors.

Goals and Outcomes

  • Use experimentation as a method for engaging with questions about writing, authorship, and technology
  • Interrogate biases, risks, and consequences of text and image generation
  • Reflect on the possibilities, limits, and ethical concerns with AI writing technologies

Materials Needed

Because of rapidly changing access to technologies, this year’s version used various generators as we were able to access them, including the following: 

  • A text generation tool: ChatGPT, GPT-3 Playground, Canva “Magic Write,” open source LLMs, etc.
  • An image generation tool: DALLE-2, Canva Text to Image, Stable Diffusion, etc.
  • Design/layout software: Canva, Adobe InDesign, etc.

Acknowledgements

I would like to acknowledge the teaching team for the 2022-2023 year, including co-instructor Andrea McKenzie and teaching assistants Rositza Georgieva, Carla Ionescu, and Ksenia Jourova. I’m also grateful for conversations with Laura Allen and James J. Brown, Jr. about how they use text generation assignments in their courses.


The Assignments 

All work from this unit is submitted and assessed as a portfolio, with emphasis on process over product to support experimentation and risk-taking, and it includes three required elements:

  • Four writing experiments
  • Final draft of children’s book
  • Reflective essay
  • Optional: any other notes, examples, or images that are relevant to demonstrating student process and engagement with the experiments and materials.

Weekly Writing Experiments

  • Exercise #1: Using ChatGPT, generate a children’s story about the topic of your choosing. You can try as many times as you like, but you should save all of your prompts and results. Because of in class activities we’re going to do, do not edit the text produced by AI yet. 
  • Exercise #2: Edit up to 20% of the story you selected from last week. Indicate in bold where you have made changes.
  • Exercise #3: Generate at least 5 potential images for your book. You can try multiple times, but you cannot edit the images yet. Submit your images with search terms.
  • Exercise #4: Layout the text and images together into a draft of your children’s book. You can lightly edit and crop images to bring together a cohesive project.

Final Project

Create a 12-page children’s book, including front and back covers, using the text and images you’ve generated. No pages should be blank, having either text, or images, or both. 

Reflective Essay

Throughout this term, the university has sent numerous updates (and warnings) advising us as teachers and students about how AI technologies should be handled in an academic setting. Because these technologies are not (yet) part of the university’s policies, concerns about the research, creation, and learning we do as a community are constantly being negotiated. This reflection asks you to think about your experience on this project as a way to consider contributing your perspective to this conversation. 

Based on your experience creating a children’s book—as well as engaging in class readings and discussions about writing, authorship, and technology—how would you recommend the university’s academic integrity policy be updated? Write a ~750 word reflection that discusses your experience in this assignment and considers how those might connect to how the university navigates this challenge. In other words, what are one or two specific suggestions for how members of our university community might navigate these technologies, either in line with or in alteration of the policy, supported by specific instances, findings, or experiences in your work this term?