Striking the right tone in a piece of writing can be really hard. How does a writer know when to be casual or serious? How much personality or humor should go into the text? Questions like these get even more challenging when writing for school because it seems everyone has their own rules for students to follow. Some teachers (especially in creative writing) want students to develop a sense of voice. Some (like Kate McKinney Maddalena) even encourage the use of the first-person point of view. Others (especially in the sciences) claim that getting a sense of the author might distract from the content. Personality, these folks say, distracts from the research. But if we actually remove the personal from our writing, what’s left? Turns out, we’re right now getting a great demonstration of the horror show that results from de-personalizing writing.
Enter ChatGPT, the Least-Personal Author Ever
Toward the end of 2022, the company OpenAI launched ChatGPT-3 and launched it as a public demonstration (for what it calls “research purposes”). A cynical view might say this public launch generated more publicity than research results, but one never knows, given the amount of data pouring into the service thanks to all the attention it’s garnered.
In short, ChatGPT is an AI that processes “natural” language responds to requests for conversation and content. Users can ask the chatbot to create text of various forms. And, perhaps most impressively, users can request stylistic changes to the results. At first glance, the system does an amazing job wrangling language and producing content. It can crank out paragraphs of smart-sounding text in no time flat.
English teachers and professional writers were terrified. Would they all be out of jobs?
The Personal Backlash
The concern over ChatGPT and its ability to create legitimate-sounding prose simply, easily, and in record time came from a worry that routine writing tasks might be co-opted by machines. And, frankly, they probably will. BuzzFeed—notorious source of clickbait articles devoid of real content—has already said it will use ChatGPT to create content later this year. In related news, BuzzFeed has laid off 12% of its human workforce. But before we start lamenting the loss of human labor, let’s think about the kind of labor being outsourced. When was the last time you felt satisfied after reading content on BuzzFeed? How much work must it take to write what they call an “article”. If we can prevent humans from creating the kind of material that floods the BuzzFeed website, I think we’re doing the world a favor. But I digress…sort of.
Many teachers worry that ChatGPT will allow students to outsource their papers. But the only writing ChatGPT excels at usually earns descriptors like “hollow” or “vapid” or “mechanical”. In other words, the chatbot creates impersonal, inhuman writing. We should expect more of students and our online content. Hell, we should expect more of humanity. If hollow, vapid, mechanical writing meets the needs of English-class writing assignments, those assignments are flawed. As John Warner articulately argues, AI “can’t kill anything worth preserving.” We need to set the bar higher for our classes. We need to emphasize the human, with or without the technological.
We’ve Been Here Already
Concerns about the destruction of writing classes are merely extensions of latent, decades-old fears. And they’re being triggered by a technology that’s little more than an iteration of tech we already use. Our computers have autocorrected spelling for over a decade. Our phones have auto-completed words and phrases for us for years. Google Docs has suggested sentence endings for nearly as long. We use algorithmic writing tools every day. The new tech seems more threatening because it’s bigger and more competent, but it’s the same basic deal.
We’ve all gotten mad at autocorrect for screwing up our spelling when we were right in the first place. And we roll our eyes at autocomplete for being ridiculous. ChatGPT is no better: Its writing is not personal. That’s super-easy for us to notice and super-hard for computers to address. Therein lies the challenge—and the value—of AI-based writing tools in English classes. We need to help students see how writing should be better than what AI creates.
Using ChatGPT in Class
This semester, each of my classes has its own blog. Students create posts each week responding to our readings and preparing for class discussions. Some posts are complex and thought-provoking. Some are short and uninspiring. But each one is personal because each one shows what that student was thinking. Each student chooses how they respond to the readings, drawing from their interests and experiences—things an AI doesn’t have.
As an experiment, I created an account for ChatGPT on each of my class blogs (Intro to Writing, Writing for Digital Spaces, and Writing Pedagogies). Each week, I ask it to create a blog post that synthesizes and responds to the readings, providing links to each one so it can process the source text. One time, I didn’t provide the source material, and you can tell—the post is complete bullshit, extrapolating a guess from the article titles but not actually engaging with the material.
Today, I asked it to create a BuzzFeed-style post. It did a great job, writing a “7 reasons you should…” post that means little, omits a lot, and makes you feel like you read something. You know, very BuzzFeed-y. But overall, its posts are awful. The AI creates sentences that sound like they’re meaningful, but you quickly realize they aren’t. The computer-generated statements lack substance. The bot just shovels words based on statistics, not actual thought.
The Nuance We Need
I think the addition of ChatGPT to conversations about writing instruction is great. Popularized, accessible language-model AI makes it easy to discuss what writers do and what makes writing good. We now have a quick and easy way to generate complete and almost-but-not-quite-good writing samples. I can use that writing in class as a baseline example and really focus on what humans can meaningfully add to writing that computers can’t. I hope this technology makes discussions of voice, style, and coherence more common in our classes and more tangible for students.