Overview of Approaches
Decision Graph / State Machine
When we developed Prince of Destruction in the early 90s, I mapped out the entire story as a gigantic flow chart (including side-quests, etc.). When I went to GDC in 1993 and spoke with other game developers, I discovered this was state of the art. Since then, all that's really changed is the tooling. In 1994, alongside our game we released our Scenario Tools and MARS engine.
In 2002, Neverwinter Nights (Bioware, 2002) shipped with a refined toolkit (Aurora) for creating custom scenarios (including dialog trees, etc.). While creating stories was not directly supported, the elements for implementing this kind of story (e.g. conversions, journal entries) were present. While undoubtedly more refined and usable than our tools, Aurora did not allow for signficantly more sophisticated game narratives, and there is no evidence of game narratives getting any more complex or dynamic since then.
The “Dungeon” as Story
This is a narrative analog of the Geography as Historical Determinism thesis put forth in Guns, Germs, and Steel (Jared Diamond, 1997) and exemplified by early TSR's D&D "adventure" modules which were little more than keyed maps the story emerged from the way players responded to the descriptions of and challenges posed by the locations they visited in the order they visited them, which was in large part determined by the map layout.
Most content editors on the market, including those with elaborate support for decision trees, are built around map editors that greatly facilitate this mode of "story" development. This was just as true for our Scenario Tools and NWN's Aurora, and is exactly how "level design" in first person shooters works.
Often when writers describe "emergent narrative" what they tend to mean is that if you stick players in a simulation of some kind and they do stuff, then the experience they have and the stuff they do is an "emergent narrative". While this is undoubtedly true for sufficiently vague understandings of "narrative", it's really not helpful, since the same could be said of any experience, e.g. trying to figure out how to assemble IKEA furniture or driving to work.
AI is really a special case of simulation in this case simulating "intelligent actors" and hoping (or asserting) that narrative will happen when you are faced with a world populated by such intelligent actors. The problem with this is that it relies first on actually being able to effectively simulate these intelligent actors (a significant feat in itself, and one not convincingly attained) and then hope these actors will somehow spontaneously do interesting or memorable things. Our world is already populated with actual intelligent actors, and yet our day-to-day experience is not riddled with interesting and memorable things.existing markdown 12/16/2014 15:03:05