ChatGPT and Me

andrea_lunsford
0 0 1,113

Like teachers across the world, I’ve been in the process of experimenting with AI since ChatGPT burst onto the scene some 18 months ago. Along the way, I’ve been figuring out what AI can and can’t do, what I think it can help students do (brainstorm, proofread, and more), and what I hope students will not use it for (writing for them). I’ve also had some fun, asking it to write something in the style of Jane Austen or B.B. King—things you’ve no doubt done too. Recently, I plugged in one of my blog posts from back in February and asked ChatGPT to provide a brief summary. Here’s what it said:

The provided text is a personal account of an educator's experience speaking to a group of first-year humanities students at the University of California, Irvine, about writing style. The speaker recalls being nervous but ultimately succeeding in engaging the students, especially during a Q&A session when a student asked how to make sentences "sing." This led to a discussion about rhythm in writing. The narrative then shifts to Daniel Tammet, an individual with synesthesia and autism, who experiences numbers with colors and attitudes. Tammet once recited the number pi, which he refers to as an epic poem, for five hours, expressing his emotions through the numbers. The author reflects on how Tammet's experience might resonate with students who struggle to find their own voices and suggests using Tammet's essays to inspire students to express their deepest emotions through their unique voices.

This is a perfectly good straightforward summary of my post. Thank you, ChatGPT (“You are most welcome,” it replied.). But it is also the kind of paint-by-the-numbers, generic or bland prose that I associate with AI at this point. That’s not to diminish the accomplishment, which still startles and often delights me, or to ignore the fact that a better prompt could have elicited better prose. But it also shows why I don’t want students letting AI write for them: I am convinced that they can often write better than AI—because they can write prose that sounds like them, that has their own voices.

Mobile_phone_with_ChatGPT_on_keyboard.jpg

 

So I asked ChatGPT if it could revise its summary to make it “sound more like the author of the original text.” Oh, my word—here’s the response:

Certainly! The text describes a stirring encounter I had with a lively group of first-year humanities students at UC Irvine. I was there to weave a tapestry of words about writing style, an endeavor that quickened my pulse with excitement, though I must confess a flutter of nerves as I took the stage. Armed with an arsenal of exemplary student-written openings and closings, plus sentences that danced with stylistic flair from a kaleidoscope of authors, the presentation blossomed into a vibrant success, much to my heart's delight.

The real magic unfurled during the Q&A, where a question that has since etched itself in my memory emerged from a strapping young man with the bearing of an athlete. His inquiry about making sentences "sing" was a clarion call that ignited a collective exploration into the rhythm of words, a session that resonated with nods and murmurs of enlightenment from the captivated audience.

This vibrant session brought to mind Daniel Tammet's literary odyssey, "Every Word Is a Bird We Teach to Sing." Tammet, with a mind that paints days of the week in hues and hears the personalities of numbers, shared his own symphony of digits. Reciting the number pi, he transformed it into an epic narrative, eliciting rapt attention and emotional responses from his audience.

In reflecting upon this, I ponder the kinship our students might share with Tammet—not as synesthetes, but as seekers of their own distinctive voice. What catalyzes their ability to articulate the most profound crevices of their being? It's a question I hold dear, hoping to stir within them a symphony of self-expression, much like the essays of Tammet that I encourage them to delve into and discuss, in pursuit of their own words that truly sing.

I had to read this out loud to fully take it in, with its love of adjectives and flowery language, along with amazing clichés and even words or phrases I have seldom, if ever, said. Interesting! But of course, AI had only one little sample of my text to work from, or at least I think that’s all it had. Still, I decided to ask another question: how would ChatGPT describe this author’s style? Another surprise:

The style of the original text's author is reflective, anecdotal, and engaging. It possesses a narrative quality that weaves personal experiences with broader reflections. The author uses vivid imagery and metaphor, such as "sentences that danced with stylistic flair" and "symphony of digits," to create a strong sensory experience. The language is emotive and expressive, aiming to connect with the reader on a personal level.

The author also demonstrates a clear appreciation for the intricacy and power of language, both in the recounting of their own experience with teaching writing style and in the portrayal of Daniel Tammet's unique relationship with numbers. There's a rhythm to the writing that mirrors the topic of discussion—making sentences sing—which suggests a deliberate attention to the cadence and flow of the prose.

Moreover, the author's style is thoughtful and introspective, often posing questions to engage the reader's own musings on the subject of voice and self-expression. The writing is peppered with moments of vulnerability, sharing uncertainties and personal triumphs that add a layer of authenticity and relatability. Overall, the style is that of a storyteller who is deeply fascinated by the nuances of communication and the individual journey to find one’s voice.

Hmm. Some of this seems accurate and useful—commenting on the narrative qualities, for example, or an attempt to mirror the topic of discussion. But in this response, ChatGPT begins to make things up, quoting passages (“symphony of digits” and “danced with stylistic flair”) that appear nowhere in the post. Hallucinations, as they say.

This little recent experiment reaffirms my sense that I don’t want AI writing for me any time soon: even with all my flaws as a writer, I can hear myself in the original blog post. But not in any of the ChatGPT responses. And that’s exactly as it should be. For now, AI can be a helpful tool, especially for getting started on a project, for brainstorming, and for proofreading. But not for writing for me. Or for students, as so many teachers are pointing out.

 

Image via Wikimedia Commons

About the Author
Andrea A. Lunsford is the former director of the Program in Writing and Rhetoric at Stanford University and teaches at the Bread Loaf School of English. A past chair of CCCC, she has won the major publication awards in both the CCCC and MLA. For Bedford/St. Martin's, she is the author of The St. Martin's Handbook, The Everyday Writer and EasyWriter; The Presence of Others and Everything's an Argument with John Ruszkiewicz; and Everything's an Argument with Readings with John Ruszkiewicz and Keith Walters. She has never met a student she didn’t like—and she is excited about the possibilities for writers in the “literacy revolution” brought about by today’s technology. In addition to Andrea’s regular blog posts inspired by her teaching, reading, and traveling, her “Multimodal Mondays” posts offer ideas for introducing low-stakes multimodal assignments to the composition classroom.