Toward a minimum viable combinatorial narrative machine

I’ve been looking at a lot of experiments in storytelling form lately, and I see three broad types:

  • Traditional linear narrative in which the narrator is control of the reader/viewer’s path
  • Choose Your Own Adventure, in which the reader/viewer chooses among a set of carefully preordained paths
  • Random access, in which the reader/viewer can skip back and forth among a lot of story material with a high degree of freedom, but without experiencing a single, cohesive plot (e.g., I Love Your Work, CLOUDS)

Deciding which form to use involves weighing author control and overall cohesiveness against reader control. But is that a necessary tradeoff?

Would it be possible, I wonder, to create a new structure, which has the sense of unity of traditional linear narrative but some of the flexibility of the others? So that, for example, each time you read/watch/play a story, you start by hitting a “shuffle” button and get the same characters and settings as the time before, but different plots, goals, and themes. Could each combination be a single, cohesive story?

I can’t imagine this being done with text. But I can imagine it being done visually, because it has been done — but only very simply, and only in Russia, and only around 1920. Soviet filmmaker Lev Kuleshov performed an experiment in which he juxtaposed the same footage of a man’s expressionless face with a bowl of soup, a girl in a coffin, and a pretty woman. Depending on the cut, the audience saw the man act hungry, sad, or lustful. In this case, the trade-off between cohesiveness and reader/viewer control was not necessary — in fact, it was the viewer’s participation, her tendency to make sense of events by interpreting them as a story, that made the two clips a meaningful unity in the first place.

So here’s my idea for Kuleshov 2.0: rather than introducing new clips every time to create new stories as Kuleshov did, start out with all the story elements you need and just rearrange each time. The relationship between the clips might have to be a little more ambiguous than in mainstream Hollywood cinema, but as Kuleshov proved, we are a story-seeking people, and given a little help we will fill in gaps where necessary in order to create a narrative.

The mechanics of this narrative machine wouldn’t be the challenge: a “shuffle” button is a simpler mechanic than the “random access” examples linked above. What would be hard is coming up with clips that are both rich and loose enough in their juxtapositions to accommodate story-making in any combination. And that gets progressively harder as you scale: If you two clips to shuffle, you have two possible stories. But if you have three clips, there are six possible stories. Four clips, twenty-four possible stories. It becomes unwieldy quickly.

That’s what I worry about when I read something like this:

The words “combinatorial explosion” have probably never come up in a video game storytelling talk as many times as they did with BioShock creator Ken Levine’s “Narrative Legos.” At the 2014 Game Developers Conference in San Francisco, Levine outlined an ambitious — if vague — idea for endlessly replayable, systems-based stories that he compares to a player-driven episode of Game of Thrones. When a slide deck offhandedly lists “build a web of nearly infinite relationship states” as a prerequisite, you’re in for something that’s at least compelling listening.

What the creator of BioShock is interested in is a game that is a lot more complicated from a mechanics standpoint than my “shuffle” button, but that’s not the thing I’m hung up on. He’s talking about a LOT of elements, resulting in “nearly infinite” relationship states.” Setting aside the question of whether each play-though is equally fun, how do you even ensure that each is sensical? How do you do narrative quality control as the number of story element combinations your machine spits out approaches infinity?

Better to start small, I should think. What’s the minimum viable implementation of a combinatorial narrative machine?

3 thoughts on “Toward a minimum viable combinatorial narrative machine

  1. Is the trick to a cohesive feel all about the transitions from scene to scene? And if so, then is the whole narrative act about creating scenes that are vague enough to “always” “work” and then using transitions to help the whole thing feel seamless and whole no matter the order? I think Ken Levine is implying some pretty impressive AI would be present to help guide his stories organically.

    Alonside creating something more or less brand new, it would be interesting to see well-known movies chopped and shuffled. Does a tight chronological narrative become total nonsense when remixed in this way?

    1. That vagueness is what I worry about. In my previous post I quoted a filmmaker who said of their interactive project “The script is the most generic version of a breakup you could ever think of. There’s no way you could apply what they’re saying to 60 different actions if it was very specific.” And that was with a story where all the possible branches converged to the same conclusion every time!

      I get the feeling that the Hollywood tropes of telling an audience something three different times and presenting neat character arcs won’t work very well when reshuffled. But you’re right, it would be an interesting experiment, especially using something with a super tight story. I think it would also be fun to take three or four clips from a free found footage source like from the Prelinger Archives and see if they can be reshuffled in coherent ways. That might be my next project.

      (I was going to mention Gods of the Prairie but didn’t know whether it was more Choose Your Own Adventure or Random Access. Maybe those categories need to be rethought.)

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s