As I talk to writing teachers about incorporating AI into their writing instruction, a recurring theme keeps crashing the conversation like a super villain in a spandex suit: the enigmatic concept of voice. How do we protect and nurture individual voices in a world where AI is poised to make us all sound like Ultron? How do we prevent a linguistic snap from erasing marginalized voices? How do we keep all of the creative universe from becoming one enormous Marvel franchise of caped conformity?
For me, the question is simpler: How can I even get my students invested in finding the power of their voices instead of just obeying the “rules” of writing to get a “good grade?” In a world where every freshman has an Insta-identity, will they care enough to craft and curate their writing as carefully as their online persona? Surely no one wants their writing to be labeled “basic,” right?
Since AI is a weapon that threatens to destroy the soul of writing, I started experimenting to find out if I could use it as a tool instead. My experiment was elementary: I asked ChatGPT to explain how voice works as a rhetorical tool. It’s the type of question my students might ask if they wanted a textbook-style answer, and it produced a pretty textbook explanation. Now, as Ultron would say, “It’s time for some mind games.”
Next, I asked ChatGPT to explain the same concept to me using the voice of Peter Elbow—again, pretty standard. Then, I asked it to teach me in the voice of Ms. Rachel, the preschool education YouTuber. Sure enough, I got a lesson sprinkled with upbeat images (literally, she told me that a vivid image is “like putting sprinkles on a cupcake”) and an affirmation that my writing voice could show readers who I am and make my message stand out.
To counteract the unicorns and rainbows, I next instructed ChatGPT to explain using the voice of Winston Churchill. As expected, there were a lot of “dear friends” and comparisons between the “realm of rhetoric and warfare.” A clear voice could “mean the difference between victory and defeat.”
Of course, I proceeded to spend way too much time “doing the voices” until I finally drew the line at Super Mario’s “Let’s-a talk about-a rhetorical voice, woohoo!” Nevertheless, the experiment provided a potential game plan. If my students could look at these examples and identify the speaker, could they also identify the tone, word choice, sentence structure, and assumed audience that created each “voice”? This could be the first step in having them start to make the rhetorical choices that could uncover their own voice—their origin story.
I know it’s a silly assignment; I told you that I asked ChatGPT to impersonate a video game plumber, so I’m obviously not trying to outsmart anyone here. I am trying to leverage what AI does well— imitation—to lead my students to find and celebrate their own unique voice, or as ChatGPT Shakespeare would put it, to “write the verses of their own poetic soliloquy.”
The experiment also opens up the possibility of larger conversations about both language and AI. Ask ChatGPT Jay-Z to talk about rhetorical voice and you won’t hear from an innovative artist who has crafted a voice, persona, and brand that is globally recognizable; you’ll get some pretty bad 90s rap lingo and some truly awful stereotyping. Why? Congratulations—you’ve just motivated students to talk about the bias of AI generated language as well as the models upon which it was trained. It’s also a chance to talk about WHY protecting diverse voices is even more important in the post-AI world.
In doing all this, will we uncover all the answers? Of course not, but maybe students will realize the importance of having these conversations with other humans and not depending on AI Iron Man to answer for them.
