Curating Simulated Storyworlds (James Ryan) – Ch 1-3

James Ryan recently graduated from UC Santa Cruz, and he was kind enough to make available his dissertation, Curating Simulated Storyworlds, for anyone to read. Of academic work coming out recently, this is one of the more interesting to the interactive fiction crowd, and I’ve already recommended it to quite a few people. I’m going to be writing about it in a few posts, since it’s long enough that I wasn’t able to read it in a single sitting.

As with other posts about academic work, I’m aiming partly to make interesting academic work on interactive narrative visible and accessible to hobbyists and people from the game industry; but I also use the opportunity to record my own thoughts and reactions to the material, and these are often based especially on the history of interactive fiction. So while Ryan’s dissertation is not primarily about text adventures, I will sometimes draw connections from his ideas to work from the text adventure community.

The basic idea: Ryan is interested in the kinds of emergent stories that can be built by Dwarf Fortress-like simulations — large, complex worlds that generate many many events over many simulated years of interaction, often with striking and memorable chains of causality. But from a narrative perspective, experiencing these worlds is not always satisfying. Sometimes they generate fascinating emergent plots. Sometimes they just seem unfocused or dull. Hence: curation. We need either a human being or a second AI system capable of extracting the good stories from the simulator and presenting those to the reader:

To understand the successes, we might ask this essential question: what is the pleasure of emergent narrative? I contend that the form works more like nonfiction than fiction—emergent stories actually happen—and this produces a peculiar aesthetics that undergirds the appeal of its successful works. What then is the pain of emergent narrative? There is a ubiquitous tendency to misconstrue the raw transpiring of a simulation (or a trace of that unfolding) as being a narrative artifact, but such material will almost always lack story structure. (xii)

This is an area that a few others have touched on; Jacob Garbe’s Dwarf Grandpa project is essentially about curating a simulated storyworld.

In essence, Ryan’s assertion at the beginning of the dissertation appears to be that the difference between good and bad emergent narrative generators is simply whether anyone is sufficiently interested to bother curating the output: so Dwarf Fortress and the Sims are good emergent narrative generators because people retell their constructs, while some academic projects are not because no one is moved to retell those. To me this did seem to miss some points about what makes generators effective, including

  • whether they use a number of systems that interlock in interesting ways (this is a somewhat handwavy description, but Tarn Adams describes the point much more effectively)
  • whether the systems account for the possibility of stakes and motivations, or whether they mostly model less interesting things
  • whether the components of the systems are polysemous or symbolically rich, thus capable of supporting additional interpretive constructions beyond what the author might have intended
  • what range of outcomes and story shapes can be achieved; the expressive range of the generator

…though it may be that Ryan will come back to those or similar points later in the dissertation.

Ryan’s approach includes an explicit, extensive discussion of the aesthetics of emergent narrative. Why are we even bothering with this, and what experiences are we attempting to achieve? What does emergent narrative make possible, and what are the problems with it?

I was very glad to see this, because I think this kind of discussion is of critical interest for people who approach these systems from an artistic perspective, and they’re often entirely omitted or at best not very thoroughly considered in academic writing on procedural narrative systems.

The dissertation is sizable, so I’m going to be talking about it in a multiple chunks here.

The early chapters of the dissertation provide a detailed definition and history of the concept of emergent narrative: what we mean by this, when the terms were first used, and what else has been written about the topic from the 90s on.

In the third chapter, Ryan begins by pulling apart what it is that we might like about nonfiction (and even the question of what nonfiction is in the first place), in order to argue that simulation-based narratives offer that same kind of pleasure. He also suggests that historical writing involves a sense-making procedure that is different from merely recording event sequences, so that some curation activity is required for nonfictional as well as fictional narratives.

Chapter three also introduces the idea of computer art brut, “art by a computer that is distinctly removed from the aesthetic sensibilities of human art”.

Then he identifies a series of ways that emergent narrative can appeal to us:

  • the actual; a sense of “true events,” even if they are simulated
  • the personal, a story being told or generated specifically for or about the listener
  • the uncanny and sometimes farcical; the dedication to strange or surreal logic that nonetheless does have some rules
  • the unauthored or co-authored; the absence of a human hand at some key point of creation, or the idea that the results are partly outside human control
  • the “uncovered”; the fact that the recipient of the story might be exploring or uncovering that story themselves, which relates to interactive stories that rely strongly on mechanics of exploration
  • the improbable; the idea that most events and sequences invented by the simulation are dull and therefore non-dull sequences, curated to the surface, are memorable because they’re unusual
  • the vast, focusing on the sheer size of generated worlds and the number of events that might be created within one
  • the ephemeral, because such worlds can be produced and destroyed with the touch of a button

I think this is a good list, though some of the items on it seem to me to belong more than others, and though it leaves off a few items that I myself would likely have included.

For instance, from my point of view, one of the values of a simulation is that it invites the reader or player over time to draw conclusions about how the world works; indeed, when I put randomized or procedural text into a game I’m building, it’s often because I want the player to get a sense for the kind of thing they should expect to see in this space, the tendency being more important than the individual instance. The generated item descriptions for abstract items in Counterfeit Monkey, the generated book names in many a text adventure library — these are pointers towards an entire physical or cultural system that can’t be adequately sketched by means of a single example. These are exactly the same cases where, if I were telling the story orally to a particular audience, I might riff and invent new examples at the moment of telling.

Meanwhile, though I like the idea of the uncovered story, it feels like this refers to a quality of how a story is presented to the player, rather than an aspect of the generative system. You can have simulation-based stories that the player does not explore, or hand-authored stories that the player does discover via archaeological gameplay — as witness nearly every classic game narrative that heavily relies on found diary pages and environmental storytelling. While it is indeed satisfying to have a simulation that leaves study-able traces of events — indices, in the sense of Clara Fernandez-Vara’s indexical storytelling — I’m not sure that the item belongs on this particular list.

The uncanny, on the other hand, interests me very much, particularly because it tends to confront us with the errors in our own thinking. If the original simulation is a simplified representation of how we think the world works, or might work, then when it kicks up uncanny or unacceptable results, these results point out the assumptions we did not bother to add in to the simulation. Testing our own assumptions via procedural art is in my view one of the most interesting things to do with the form.

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

%d bloggers like this: