I am fascinated the whole phenomenon of generative AI as a fiction-writing machine. Not because I think it’s a good idea (it’s a terrible idea), but because of what it tells us about what we think the work of writing fiction is, and also what we think stories do.

In order to even consider automating novel-writing, you’d have to believe that labour of writing a novel is split into three distinct parts that can safely be teased apart: coming up with the setting, characters, plot, and themes and documenting all those (the creative part), painstakingly and mindlessly lining up words down in the word mines for months at a time, and then coming up for air to do some light editing. And voila! A novel is born. You have to believe that the only creative part, the only parts where decisions are being made, are first and last parts, the planning and the editing. You would have to embrace the idea that there is no creative work happening in the actual composition of a novel. It’s just arranging deck chairs, the brain work is done. Like a novel is a film shot on doorbell cams, and the main thing is to plan the action well and edit the footage thoroughly.

For people who haven’t written anything, this idea might make a lot of sense. An architect gets credit for a flashy building even though they didn’t do any of the construction. Jony Ive didn’t physically construct the iphone. For people who don’t write, these might be relevant comparators. An outline is a story in miniature. If what you value is the big ideas, the execution wouldn’t hold as much value.

The one terrific thing about this AI slop situation is how the product of those ideas in action shows us how absolutely, eye-wateringly wrong those ideas are. Without that contrast, I’m not sure I would ever have fully appreciated the various scales of creative decision-making that goes into writing a novel. Most of them I had taken for granted.

Characters, settings plots? Sure, these are important, but they don’t exist outside their implementation. If 12 writers wrote stories using the same outline, you’d get 12 different stories. The outline is a coat closet, not the coats. There are many levels of creative decision-making that happens in the process of constructing a novel, and outlining is just the broadest strokes. While there are outliners and pantsers, every outliner is also a pantser, because every time you write a scene you are discovering something. For a pantser it might be the next step of the plot, but for an outliner, it could be the intimate details of a setting, the way a line of dialogue lands on another character, how that plot element actually feels in action. What a character notices, or hears, or smells along the way. What they’re wearing, how comfortable they are, what the neighbours see. The story is in the outline, but it’s also in those tiny details, each of which is a decision. The simile you reach for to describe how a character sounds, or moves, or feels: those are all critical to the shape of the story and profoundly creative choices. If you outsource all of that to a machine, what’s on your pages isn’t your creativity. It’s just stolen creativity based on the work of others.

I had never really thought about the process of putting similes and metaphors together, to be honest. I know enough not to use the clichés, but I hadn’t given the process much thought until I saw how genAI does it. They’re empty at best, and far too literal at worst. Machines can’t tie a relevant, visceral human experience into a beat of a story. It doesn’t know how to pick which visceral human experience applies in that story moment. How could it? Without reaching for other people’s metaphors, how could it generate something both unique and authentic?

Credit here to Frankie’s Shelf, who illustrates this so beautifully by painstakingly dismembering Shy Girl by Mia Ballard. It’s so weird to me that I learned to become more aware of these elements of the creative process by listening to every last second of this 2 hour and 40 minute video.

The similes and metaphors really underscore the point: even the hard labour down in the word mines is profoundly creative work, full of decisions word by word. It’s not rote, it’s not just lining up text to get through the outline. The work at the line level is as creative as the work at the outline level.

The other critical piece I learned from genAI fiction is that we aren’t just orchestrating a rising and falling plot line (or set of plotlines). We are orchestrating an emotional journey, and it has its own pacing that can feel uncanny or inauthentic if slightly off. When genAI produces a novel scene by scene, it delivers a scattered, incoherent emotional throughline that hits an emotion at the same volume one scene at a time. Humans don’t do that. I feel like I’ve taken the emotional journey construction for granted. It’s very natural for a human, but would have to be explained to a machine, and I’m not sure any of us even know how to explain it to a machine.

Who was it who allegedly said, “I hate writing, I love having written”? That feels like the tradition genAI-using “writers” are falling into, but to an extreme that overshoots it by 5000 kms. The bird-by-bird writing of a sentence, a paragraph, a page is the creative work of writing a novel, that’s the whole thing right there. If you don’t want to write novels then don’t write novels, you know?