Today’s post is by author Audrey Kalman (@audreykalman).
On April 7, frustrated with my progress on a third full revision and still dancing around my novel’s central story question, I typed the following into my ChatGPT account.
Me: I am struggling with defining the one central thing my novel (currently in draft form) is about. Can you help me figure this out?
In roughly a millisecond, ChatGPT responded:
Of course, I’d be happy to help! Can you provide me with some more information about your novel? What genre is it, what is the general plot, and who are the main characters? This information will help me understand your story better and assist you in defining the central theme or idea.
I provided a short, rough synopsis. Seconds later, ChatGPT returned several paragraphs summarizing what I had provided and offering a possible theme: “The struggle to reconcile tangible and intangible aspects of reality, and the transformative power of belief in the unknown.”
Okay, that’s actually two themes.
Were either of them really what I’m going for? I’m not yet sure. But in a few minutes, ChatGPT had cleared the fog that often results from being too close to your own work and gave me a starting place for refinements. As I write this, a little more than a month after my first interaction with ChatGPT, I have a revised outline, an entire new story line for one of the characters, and a clear direction for my revision.
The Chatbot is in
I’ve checked in regularly with ChatGPT since that first session. I pose writing questions ranging from the general (character development) to the specific (developing a scene about a natural disaster). Occasionally I venture into the more emotional side of the creative process:
Me: I’ve been feeling frustrated at how long my revision is taking. I’m spending so much time planning when I really want to be writing. Do you have any suggestions for tempering my impatience?
ChatGPT: It’s completely understandable to feel frustrated and impatient during the revision process. Here are a few suggestions to help temper your impatience.
The chatbot’s suggestions for dealing with frustration are similar to those I might dispense to coaching clients: embrace the process, set realistic goals, take breaks, and practice self care. Nothing too original, but these interactions advance my work significantly by sparking further ideas.
Using ChatGPT this way feels qualitatively different than my earlier experiments with everything from research, compiling content, and generating comp titles to asking for plot suggestions that better reflect a character’s story arc.
This feels like a coaching relationship.
Having both provided and received coaching, I appreciate the back-and-forth, generative quality of the interactions. That’s what excites me about teaming up with a generative AI system. Talking with ChatGPT the way I might with my long-time writing partner or a writing coach led me somewhere I never would have gotten on my own.
Can a generative AI be an effective writing coach?
ChatGPT possesses what I estimate to be about 70 percent of the skills that make an effective coach, which makes me a little nervous as someone who still offers human coaching. It’s good at:
- listening and reflecting
- making guided suggestions based on stored knowledge
- responding without an agenda—a place I’ve had to work hard to arrive at as a human coach so I can offer nonjudgmental, non-prescriptive support; without its own experiences to share, ChatGPT is objectiveless by definition.
In some cases, a generative AI is better than a human. It doesn’t sleep, take a lunch break, or stifle a yawn in the middle of your rambling explanation of a writing problem.
However, the 30 percent of the competencies that it lacks represent the real juice of human-to-human coaching relationships, making a strong case for engaging a human coach, either exclusively or in addition to an AI-based one. That 30 percent includes the following skills.
1. Intuition
Often, human coaches can intuit what writers should focus on, regardless of the issues they bring to the coaching session, and gently guide them in a different direction. ChatGPT may not tell the hard truths about your situation, such as: “At this stage, worrying about story continuity might not serve you. Maybe you should focus instead on how readers will form an emotional connection with the characters.”
2. Reading between the lines
The ability to read between the lines, to interpret subtle language cues (especially body language in face-to-face interaction), and to build on the foundation of an ongoing relationship remain—for now—in the human realm. A human listener can respond based on information that goes beyond the specifics of a prompt.
3. Human pacing and cadence
Humans have been talking to one another for hundreds of thousands of years. As frustrating as our communication can be, there is comfort in its human-paced rhythm. ChatGPT, by contrast, can respond within seconds or even fractions of a second, a rate our conscious minds can’t match. Human/AI conversations tend to be lopsided as the human pauses to absorb an answer and formulate a reply.
Of course, future enhancements could make ChatGPT’s response cadence more human. Think, for example, what might happen when ChatGPT gains a human-like voice. Will it mimic human communication tics such as formulating an answer before you have finished speaking, or cut you off mid-sentence?
4. Experiential empathy
Many writers enjoy writing in the company of others or gathering to commiserate about writerly dilemmas. We take pleasure in these activities largely because of the feeling that we share common challenges. ChatGPT’s empathy is a form of mimicry based solely on gathered data.
This, too may evolve. Although I remain aware for now that the empathy is artificial, the human tendency to anthropomorphize technology has been around for a long time. There’s a body of research about emotional attachment to artificial intelligences, and no reason to think this won’t apply to chatbots like ChatGPT.
5. Handcrafting and physicality
The introduction of digital music was anticipated to doom the vinyl record album. But interest in vinyl has surged because interacting with music through a physical medium is a qualitatively different experience than interacting digitally. Similarly, literary artists and their appreciators are offering their creative talents in physical form in ways that a generative AI cannot reproduce.
6. Embodiment
We are more than our thoughts. Since humans first began philosophizing, we have embraced the idea of mind/body/spirit. ChatGPT has no body, and by its own admission, no spirit. (When asked directly about the topic, it responds, “As an AI language model, I don’t have feelings or opinions.”) AI’s access to facts may be nearly limitless, but its knowledge of the physical and spiritual dimensions come to us filtered through intellect.
Should you explore using ChatGPT as a writing coach?
I believe you’ll be better served to explore ChatGPT’s potential to augment human creativity than to lament what it may take away from writers, but your answer will depend on your willingness to experiment and how comfortable you are with both technology and uncertainty. To benefit, you’ll be well served by an ability to ask the right questions in the right way, perhaps honed by years of search-engine research, so you can craft prompts that will elicit valuable responses.
Your answer also depends on how you feel about the larger issues around generative AI such as what value humans bring to a world in which artificial intelligence is finding its way into more human endeavors.
We’re being called on once again to redefine ourselves as humans, examine our most deeply held beliefs and values, and emerge with an understanding of how technology can work for us rather than the other way around. Who better to do this than a group of people—writers—known for imagination, communication chops, and an ability to dream new worlds into existence?
Audrey Kalman writes fiction with a dark edge, often about what goes awry when human connection is missing from our lives. She is the author of three novels: What Remains Unsaid and Dance of Souls (both available on Amazon) and The Last Storyteller, available on Kindle’s Vella platform. Her fiction collection Tiny Shoes Dancing and Other Stories was shortlisted for the Rubery International Book Award. She co-created the Birth Your Truest Story writers’ community and workshop series (next sessions begin in January 2023). She offers editing and coaching for writers and is at work on another novel. Connect on Twitter, Facebook, or via her website.
Thank you. Very interesting. I teach novel writing online and this article has proved useful.
Audrey, this is quite helpful in understanding more about ChatGPT. I’ve been very confused about it. I’m hearing that many writers think the software is a better writer and are substituting themselves with what they think is a higher intelligence. One thing that is ever present for me is the personal creative process. The muse. I don’t see how any software can replace the excitement and discovery of hearing the muse speak from my own explorations, my research path, my life experiences and insights into the storytelling as it evolves. It’s a thrilling ride and why I write. The muse creates soul writing for the most authentic fiction. I think Joseph Campbell said it best when speaking about hearing metaphorically instead of concretely:
“Anyone writing a creative work knows that you open, you yield yourself, and the book talks to you and builds itself. To a certain extent, you become the carrier of something that is given to you from what has been called the Muses, or, in biblical language, ‘God.’ This is no fancy, it is a fact. Since the inspiration comes from the unconscious …” The Power of Myth, page 71 in the paperback.
Honestly, I don’t see why we need or want ChatGPT when everyone has their own spiritual power and depth to write from. We can trust our own human spirit to speak to us, yield to the inspiration of our inner muse, and be open to the book to build itself, from ourselves.
For me, too, the joy of writing is the process itself, what you call the “thrilling ride.” No entity, whether biological or digital, can have that experience in my place!
However, when it comes to the next step–translating what comes through me from the unconscious into coherent language that other humans can understand–I believe tools like ChatGPT (or trusted friends, or editors, or coaches, etc.) can be extremely helpful. At least, they’ve proven to be so for me.
Audrey,
This is such an interesting relationship you are exploring with Chat GPT. I love the idea of creative people using it as support by feeding very specific questions. My own foray only resulted in a competently written but very bad poem. Thanks!
It’s good to know that writing bad poetry is not a uniquely human skill! I hope my description here has revealed why I’ve been so captivated by ChatGPT for the last few months.
That is a really interesting use of ChatGPT. Thanks for sharing.
Hi Audrey,
I really like your approach to AI. It’s another tool. Experiment. Learn.
Probably every tool since the fire and the wheel has been put to good use or ill along the way. AI will be no different. Its potential dangers are widely discussed — and without guardrails, AI poses some serious threats. Meanwhile, doesn’t it make sense to learn about it so that we can be informed and use it to make our lives better and more interesting?
You’re doing just that, and I applaud you for it. You’ve already sussed out several ways it can benefit you and your career, as well as its limitations.
I appreciate your sharing what you’ve learned so far with the rest of us.
I agree–the head-in-the-sand approach won’t make it go away and instead just makes one ill-prepared for whatever developments occur. I’d rather be informed and educated to the best of my abilities.
I’m writing a book about an event that occurred in WWII. Because I want the book to stand up to academic rigor and because I want it to sell well, I’m focused on accuracy and citing my work. I use AI as a research assistant which has been helpful in finding resources I hadn’t found. It did return one hallucination. I’m also using it to help fine-tune my author’s platform marketing plan and will probably use it to assist with my newsletter. But I want the writing to be my own.
That sounds very similar to the approach I’m taking. Good luck with the book!
This was great, and I wholeheartedly agree!
I tried ChaGPT and found it brilliant. More recently, however, it’s grown censorious. I write speculative fiction on the fine line of sci-fi, where the going can get rough and edgy. CGPT nailed me for language, references, and behavior. Bad, Jeff. Well, the boorish machine-language reader, CGPT, missed the whole point. So I tried BARD (Google’s machine-language program), and I told it not to censor me. It didn’t. Its responses were clear, detailed, and helpful.
I, too, have found ChatGPT to be tone-deaf on many occasions. This speaks to one of the challenges with all of these large-language-based chat bots–lack of transparency about the rules they follow behind the scenes. It leaves us as users to experiment and try to figure it out. Happy experimenting!