Narrative and Computational Text Analysis was one of the intensives I was most excited for, and it certainly taught me a lot. I could probably write entire blogposts on every section of the course, but I’ll stick to things I consider to be relevant to my final project:
Of the pre-intensive readings, what particularly stayed with me was Nothing Breaks Like an AI Heart – both in the sense of how well it creates narrative links between the human and the generated content, and in what I wish it had expanded on. The side-arcs are possible choices, but they never go very far (which makes sense as the main path aligns with the author’s main points, but it’s still repetitive and unsatisfying as a reader) and, similarly, clicking through the text-generated options becomes less fun when you realize it doesn’t actually do anything. (It’s still a great text. Highly recommend.)
The structure worked for this particular text, but I’m enamoured with the idea of creating a narrative in which tiny, inconsequential choices at the beginning do matter in the end, even if readers themselves aren’t aware. A narrative butterfly effect that only rereading will reveal.
That idea was compounded by the post-intensive work. For the essay component, I chose to write about the narrative structure of a Choose Your Own Adventure story generated by ChatGPT and its strengths and weaknesses. (It has a lot of weaknesses). One lightbulb moment was this text about gamebook structure, which revealed a lot of cool little narrative tricks that I’d never encountered before, as well as ones that I now remember from reading CYOA type stories as a kid (like creating bridge sections between a choice and an outcome to keep readers from remembering the section numbers that kill off the character and avoiding them). That won’t necessarily be relevant for a digital interactive narrative, as readers won’t be able to see where section links lead them, but it’s a narrative uniqueness that I want to dig deeper into to see what I can use.
(Interestingly also, the lecturers for the course simply assumed that everyone already had an OpenAI account, and I earned a strange look when I started the exercise by registering for it. Is it really the norm that everyone just uses ChatGPT?)
My main criticism of the ChatGPT story is that it gives you a lot of information that, to a human reader used to human writing, implies that all of it will become relevant again at some point in the narrative, only for it to never be mentioned again. In some ways, that’s due to ChatGPT’s limited memory, but it’s now awakened my ambition to create a branching narrative in which the choices given in the beginning do matter at the end – even the little ones that seem inconsequential – and in which narrative ideas established early on do get referenced again before it’s over.
Looking back on everything I encountered during the intensive now, I’ve come to the conclusion – besides ChatGPT being very good at ‘fake it ‘till you make it’ – that gamebooks are a lot more sophisticated than I thought. I’m excited for my final project now, and I can’t wait to start exploring all the narrative possibilities an interactive form affords.
Leave a Reply