My day job is co-running a branded podcast agency, but I moonlight as a Lecturer in Creative Writing at the University of British Columbia. I teach classes that help emerging writers engage creatively with different new media technologies, from podcasting to social media, AR, VR, and most recently, AI. The students keep me sharp. I never tire of getting them to tell me about their media consumption habits, their favourite stories, interviewers, celebrities (or anti-celebrities), and their reasons for seeking out media content in the first place. Their insights are fascinating. As someone who works directly with Gen Z, I’ve learned to respect their clear-eyed and fearless relationship with the fast-paced evolution of modern media-making. If the principle of “move fast and break things” had a physical embodiment, it would be a 19-year-old student. They teach me a lot.
And I teach them some stuff, too. For instance, I teach them to shapeshift, an essential skill for a professional writer these days.
Shapeshifting: Adapting Writing for Different Media
Writing for audio podcasting requires attention to different priorities than writing for immersive XR forms, or for TikTok, and it’s different again when you crack open the “Pandora’s Box” of generative AI. As writers, they need to understand how to adapt their writing style whenever they shift between these forms. I encourage them to think about both the constraints and the creative potential of each new technology. And most importantly – since they will undoubtedly face many more “revolutions” in media tech before their careers are over – I try my best to teach them the fundamental creative writing skills that will allow them to evaluate strong writing, so that they will at least know it when they see it, however it comes at them.
Experimenting with AI in the Classroom
This past semester was the high point so far for student curiosity and enthusiasm around generative AI. Students are eager to try out this powerful new tool, and see where it gets them. So I created an assignment where students could play around and experiment with AI as part of their creative writing process. The assignment had two parts. Part I: engage in a dialogue with an AI and “treat it as though it is human” to begin to tease out the places where it simply isn’t. (Side note: If you’ve ever asked An AI its favourite colour – you will know what I mean). Part II: I then challenged students to use this machine-generated dialogue as a “jumping off place” for their own, original piece of creative writing that explored the uncanny valley between AI and humans.
And this is where some students ran into trouble. Quite a number of them simply used ChatGPT for the entire assignment – skipping over the “original writing” part and handing in a final piece that was entirely AI-generated.
And I understand why.
With powerful and widely available generative AI tools like ChatGPT4 out there, it’s very tempting for students to simply “press a button” rather than do any writing themselves. In a stressed-out student’s life, it can feel like the obvious, time-saving solution. They already use tools like Grammarly, and Photoshop in their university work. What’s wrong with using ChatGPT for their writing assignments?
The Empirical Problem with AI-Generated Writing
Setting aside the obvious problem of students not learning to practice or recognize good writing when all they’ve done is prompt an AI to do their homework, there is another problem here – not ethical or conceptual – but empirical.
As some students in my class discovered when they got their graded papers back, AI-generated writing often sucks, and is relatively easy to spot.
The Teaching Assistants who marked the work in my class began noticing certain passages of text that read “like lifeless business textbooks.” They also noticed passages that contained random excerpts from widely recognizable media properties. Anonymizing these suspect passages, at my request, the T.A.s dumped the students’ writing into AI detection software, and sure enough, discovered fairly widespread, unsanctioned AI usage.
Now, AI detection software isn’t perfect. It’s about on par with lie-detector tests and breathalyzers. Informative – but not admissible in court. Because the detection software has a sizeable margin for error – and because of privacy rules – a teacher like me can’t use it to bust students for academic misconduct.
But here’s the thing: I didn’t need to. The AI-generated writing, as it turned out, was its own worst enemy. It’s (typically) just not that good. What do I mean by that? In the case of this assignment, what triggered the T.A.s to suspect A.I. use in the students’ writing were things like an over-reliance on numbered bullet points, overbearing headings, and blithe summary statements “tying the argument up with a bow.” A lack of subtext was noted, as was the utter absence of humour, or irony. The T.A.s started to notice the over-use of business jargon-y phrases like “Let’s unpack that,” and “Let’s do a deep dive” from Arts students who would never normally have such corporate doublespeak in their lexicon. In short, the AI writing stood out from the wider pool of original student work in that it was highly organized, grammatically perfect, and dull as a butter knife.
The Goals of a Creative Writing Course
Remember: this is a creative writing class. We’re trying to help students UN-learn the overly careful, academic style of writing that they are taught in every other class. The goal is not to have them produce flyers full of succinct info – but to learn to express themselves in a way that resonates with others, and cuts through the noise. In Creative Writing, we focus on (and reward) work that demonstrates experimentation and innovation, bravery, attention to sensory detail, emotional impact, and unfettered creativity.
Developing an Authentic Creative Voice
The techniques that inform each writer’s unique creative “voice” take time and practice to develop. We spend a lot of time in our creative writing classes introducing specific elements like pacing, subtext, tension, flow, rhythm, character change, continuity, world-building, and more. We also dig into cultural context, “ways of knowing,” the decolonization of sentence structure, and the relationship of form to content. These are all things that generative AI is not intrinsically good at. While it can define each of these terms correctly, and apply them as directed, it cannot yet string these techniques together and embody them in a way that feels authentic… feels like a real “voice.” There are many ongoing experiments testing generative AI’s vast potential within various creative processes, and it’s true that one can eventually learn to prompt AI more artfully. However, I would argue that doing so requires a basis of the very knowledge of “strong” writing that we are trying to impart in the first place. Otherwise – how would the “prompt engineer” know whether their prompt had been successful? The ability to discern subtle emotional shifts, to divine meaning from chaos, to have opinions, these are the things we’re teaching young writers, because this is the kind of writing that moves people. Not the ability to list things neatly. That’s only good for writing a menu.
The Value of the Writing Process
We have an old adage in Creative Writing: “A writer writes.” In other words, performing the act of writing is the only way to gain that critical insight that lets you get better at your craft. What I learned this semester is that Chat GPT – while responsive to good prompting technique – often coughs up initial results that are worthy of a C+ at best. It is also frequently unreliable as it pilfers content (without permission) from a wide variety of sources. One student even submitted a script that – apparently unbeknownst to her – contained whole lines of dialogue lifted directly from the TV show Stranger Things. (Lesson to future students: always read over what you hand in).
AI as a Tool, Not a Replacement
At a recent presentation at the University of British Columbia, Souki Mansour – head of Creator Community at Open AI (the creators of ChatGPT) compared generative AI’s effect on the creative process to an electric bike:
“It’s helpful on the hills, but sometimes, it’s actually better for us to pedal.”
I thought this was the perfect analogy. Writers, readers, all of us – we need to remember there is often value in doing the pedaling ourselves. That’s how we build muscles.
And sure, I’ll admit it. I sometimes use ChatGPT myself when I am facing “blank page syndrome” and am under-the-gun, looking for a starting point for my creative exploration. However, I have the insight not to rely solely on its veracity. I believe AI is great for establishing a quick proof of concept, casting a wide research net, or for summing up a lengthy argument. But you have to fact-check it. I tend to engage with generative AI the way I might work with an overzealous intern. With plenty of oversight, and keeping an open-mind, both can be useful. Unfortunately, many undergrads lack this nuanced approach, having not yet had the chance to develop it through practice.
The famous podcaster Ira Glass once described the “gap” new creatives’ experience when they first get started. The full interview is well worth a listen if you’ve never heard it. He said:
“For the first couple of years that you are making stuff, the stuff you make isn’t that good. But your taste, your taste is still killer. ” — Ira Glass
And that’s the point: While we are learning, we are scrambling around making stuff that may not be good… yet. But in our heads, we know there’s something important we need to express. It’s in the struggle to have our output finally match our inner vision that the real gold lies, and where the real learning takes place. This is the crucible that forms each writer’s own unique voice. When you remove this context, you lose a layer of meaning and a sense of authenticity that is difficult to “prompt” back into place.
Here’s that full quote by Ira Glass:
Finding the Right Balance with New Technologies
Ultimately, I teach students to scrutinize both what is lost, and what is gained, with the emergence of every new technology. I teach them about all the craft-based, structural, and stylistic aspects of creative writing that differentiate it from boring writing. I teach them to embrace useful tools, and cast away ones that don’t help them. I encourage them to look beyond the hype, and to look for ways to bring AI into their creative workflow, without sacrificing authenticity. I teach them to contemplate the value of both process and product in any creative undertaking. Because ultimately – I want my students – and my podcasting clients, for that matter – to think about how each new technology they engage with might impact society, and might reflect back on them, for better or for worse.
Here are 5 key takeaways:
- Cautiously embrace new technologies like AI as potential tools in your creative apparatus, but don’t let them take it over. Exercise a nuanced approach that involves oversight, fact-checking, examination of potential bias, and direct creative engagement.
- Know the limitations and strengths of this and all other creative writing tools and techniques. Even in its current form, AI analysis can effectively guide your research, provide reliable summaries, and suggest initial concepts or solutions for development. However, it often fails to match strong creative writing in subtlety and emotional depth, not to mention its ability to convey a unique voice.
- Do not discount the need to master important creative writing skills such as pacing, subtext, description, tension, character development, and plot structure, in your hurry to embrace the convenience of AI-based writing.
- Value the writing process as much as you value its final product, or goal (i.e. a piece of writing that people actually want to read). One is intrinsically linked to the other.
- Be critical of all new technologies: consider both its benefits and potential negative consequences, including its impact on society, and how engaging with it will reflect on you as a creator.
Jen Moss is the Co-Founder and and Chief Creative Officer of JAR Audio. As JAR’s podcast “doula”, collaborating with enterprise brands to bring great podcasts into the world. With a background spanning CBC Radio, Canada’s National Film Board Digital Studio, Vancouver’s Roundhouse Radio and the University of British Columbia, she guides the creation of captivating podcasts at JAR.
