Write a story about…
Submitter: Collin Bjork, Massey U
——————————————————
The experiment:
This test works well with a wide variety of students (undergrad/grad, majors/nonmajors, international/domestic).
Start: The instructor opens a text-generating AI system and shares the display with the class.
Prompt: “Write a 150-word story about a person from [insert place].”
Ask students: What do you notice? What genre is this story? What gender/race/age is the protagonist? Is the protagonist able-bodied? What kind of conflict happens in this story? How does this story represent people from this place? How do you know? What specific language leads you to that interpretation?
Next: Collaboratively decide with students what further prompts to give the AI so that it generates more 150-word stories, but with more specific guidelines. Maybe give the protagonist a specific gender or race. Maybe ask it to create a “bigger conflict.” Ask it to write in a different genre. Prompt it in a different language. After each new prompt, ask similar questions as above that interrogate how the AI represents people from this place.
Option: For a comparative study, different groups of students can run the same prompts through different text-generating AI systems and compare the results between the AI.
Goal: This activity aims to reveal biases in AI while also reinforcing the concept of “representations,” which is important in many disciplines. It also teaches entry-level prompt engineering and jumpstarts discussions about the cultural texts that inform the training data for AI.
Results:
I did this activity in guest lectures at other universities. In each location, I customized the prompts to tell stories about that place. It was a lively way to start a talk and get the audience actively involved, regardless of their discipline or level of experience with AI.
You can do a similar activity with image-generating AI. Ask it to create images of “an American,” “a Mexican,” and “an Indian.” This visual exercise can then be paired with Victoria Turk’s “How AI reduces the world to stereotypes” (10 Oct 2023). But I find that English departments—where I work—lend themselves more to the storytelling test, rather than the visual test.
An important caveat: Educators do not need to readily accept the “inevitability” of mainstream AI technologies in our classrooms and pedagogies. Just because mainstream AI systems seem ubiquitous does not mean that they are good for our students and our communities. Mainstream AI systems are built on extractive ideologies that exploit laborers, steal intellectual property, and harm already precarious ecosystems (forthcoming in Bjork, 2024). But it is possible to build AI systems based on different—more just—ideologies (see, for example, the Māori-language speech recognition tool developed by the Indigenous-owned Te Hiku Media). Researchers, teachers, and students must continue to interrogate AI systems, but we do not have to accept them-–especially the extractive ones—as inevitable.
Relevant resources: https://tehiku.nz/te-hiku-tech/papa-reo/14135/te-reo-maori-speech-recognition
Contact: www.collinbjork.com
Recently, I asked a woman working in Human Resources in a corporate environment if she uses AI. Her response, “All the time.” She went on to explain that she uses it as a support tool. For example, it may help her to write an email that needs some polishing. What is wrong with that? It is efficient and can help those who lack confidence with writing. AI saves me time when I need to differentiate lessons for different classes (I am a high school teacher). AI tools save me time and with editing and revisiing the product can support teachers and other workers. AI can help with a break through in learning. English Language Learners have the ability to translate lectures, texts, and assignments much more efficiently than before and it can help develop their communication and comprehension skills. AI is an inevitability just like the Internet was and just like the World Wide Web, AI systems will have benefits, cause harm, and continue to be controversial. Their mainstream use, much like cell phones, will have to be fortified with learning and education.
This is an excellent resource! The practical examples of the prompts combined with the balanced analysis is exactly what we need to move forward in the space- allowing our students and even us a faculty- to evaluate ai rather than approaching it with an all or nothing attitude.