Curating Simulated Storyworlds (James Ryan) – Ch 6f

Screen Shot 2019-04-06 at 1.14.45 PM.png

This is the third of several posts about James Ryan’s dissertation, Curating Simulated Storyworlds. We are now reading chapters 6 and following, in which Ryan describes his own projects in the curated emergent narrative space.

After the first five chapters, this piece becomes considerably more narrative in its own structure: Ryan is (consciously) telling the story of his own artistic development and practice, and the particular works to which it gave rise.

The projects described include

  • Lineage, a genealogy generator
  • World, which added terrain and exploration to the generations of created characters; murders and blood feuds and mutinies; and (most ambitiously) a concept of language that abstractly represents how dissimilar two speakers’ language happens to be; here Ryan’s linguistics background really comes to the fore
  • Diol/Diel/Dial, an encyclopedic recording of the events in World, one form of which can be explored online
  • Talk of the Towna simulation of an American farming town established in 1839 and run forward until 1979. As the name might suggest, this project pays particular attention to what characters know and what knowledge is transmitted from one to another.
    • Talk of the Town includes items such as procedurally generated gravestones that characters can read to update their beliefs about past people who lived in this space
    • Some details about a person are considered more salient, and therefore NPCs are more likely to notice and remember those attributes. A character’s belief may form because they observe evidence; because they hear something reported to them (truly or falsely); or because they transfer information or accidentally confabulate based on related information they already have in mind
    • Repeating a belief aloud makes the person more committed to what he or she has said, with the fascinating result that speaking a lie over and over can make a character believe the lie
    • However, as Ryan himself admits, the knowledge subject to all this simulation is itself not very prone to generating narrative intrigue, since characters are mostly remembering and talking about such things as the hair and eye color of other characters
  • Bad News, a human-curated performance based on the town generated by Talk of the Town, in which Ryan would extract storyline information from the simulation and it would be presented to the single human interactive audience member via Ryan’s collaborator Ben Samuel, who is also an actor.
    • Ryan describes the concept thus: the storyworld for each performance is procedurally generated prior to the start of gameplay. During gameplay, a player explores this storyworld and converses with its denizens in an effort to find and notify the next of kin of an unidentified deceased character.
    • Bad News as an experience contains elements of immersive theatre, tabletop roleplay, interactive fiction (descriptions are displayed to the player via a tablet), and improv. I couldn’t really do it justice with a summary, so the curious should definitely seek out the description in the dissertation.
    • Bad News also observes a contract of care, a way of treating the participant in the experience to try to make sure they are looked after emotionally and given a chance to transition back out of a potentially intense experience
  • Hennepin, an iteration of Talk of the Town that does finer-grained details and therefore can take a day or two to run; and which attempts to make computational the story sifting work Ryan did personally during runs of Bad News
  • Sheldon County, a generated podcast making use of the output of Hennepin

In the latter projects, the simulations are inflected not only by technical and design considerations, but also by Ryan’s interest in day to day life in the Minnesota area.

Talk of the Town handles character-character romance in an abstract, probabilistic way based on the personalities of both characters, their ages, and the amount of time they wind up spending together; the discussion of this system is quite detailed.

*

Several design patterns and techniques arise from Ryan’s observations. In previous chapters, he had already written about systems for tracking causality. In these, he introduces ideas such as “apophenia hacking,” namely allowing a human curator to build on (but not contradict) the stories produced by a simulation, so that the human is seeing and expanding on ideas that are more nuanced than the simulation actually supports.

This is something I have certainly done in my own work, especially Annals of the Parrigues, in which material produced by the prose generator suggested ideas to me that I then folded into future generations.

Another strategy, implied in earlier chapters but explored more deeply in these chapters, is the “overgenerate and curate” method, which instructs the creator to run the simulator to create more material than will actually be needed, then winnow the results.

*

In addition to his own work, Ryan mentions and in some cases records at length notable projects by past researchers, which include:

A language acquisition simulator by Sheldon Klein, which Ryan describes thus: In Klein’s unnamed language simulation, developed approximately between 1964 and 1974, agents in abstract speech communities converse with one another over the course of a few simulated decades. Each speaker maintains a both a generation grammar and a recognition grammar… Klein’s agents are modeled according to a number of concerns—“age, sex, village, clan, religion, household, marital state, work groups, and social status” [599, p. 8]—and a loose social simulation determines who will interact with whom, in a way that depends on these attributes. As the simulation proceeds, agents die and new ones are born from existing ones, leading to language acquisition.

PsychSim and Thespian, two previous projects in which character knowledge is tracked, which were influential to Talk of the Town.

Ian Horswill’s MKULTRA, which I’ve encountered elsewhere but which is re-explained here; it might be of particular interest to parser enthusiasts because of the parsing method it uses: it will suggest autocompletions to whatever the player has started typing, and refuse to accept any input that it isn’t going to be able to understand. In that respect, it’s reminiscent of some of Jon Ingold’s work on interactive parsing, or An Earth Turning Slowly.

Dwarf Fortress also gets fairly extensive coverage here, with excerpts from personal communication with Tarn Adams.

*

Ryan’s dissertation concludes by suggesting some additional directions for work in this space, including an experience that provides the player with simulation output and certain means for sifting it, and lets them find the stories that interest them. (An approach that made me think of 18 Cadence writ very much larger.)

Leave a Reply

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

WordPress.com Logo

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

Google photo

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

Twitter picture

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

Facebook photo

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

Connecting to %s