Notion: Story Management in RPGs and IN

In thinking about how we might make RPG story management more dynamic, it would probably help to review how it's done normally. The following view essentially comes from D&D's DMG, though it's been shaped a lot by my thinking on the problem of interactive narrative (IN).

An adventure is composed of encounters. Encounters are some short segment of interaction, usually phrased in if/then terms. For example, if the PCs enter this room, this monster will attack them. Or, if the PCs agree to lay down their weapons, the king will listen to their request; otherwise his guards will attack.

Encounters are really simply the story/world/GM response to player actions. So what determines possible player actions? The game rules and GM's description of the world provides affordances for PC action. That is, players know what sorts of actions they can perform based on their skills and powers, including the likelihood of success. The game rules and story context so far can provide constraints and guidance too--some actions are not possible in the world, would be out of character, or would be irrelevant/nonsensical to the story.

In terms of my IN poetics, the events internal to an encounter are world-level events. I call this set of possible world-level actions verbs. As already mentioned, the game rules provides the set of possible verbs. An encounter needs a setting (location, props, etc.) and characters (PCS and NPCs). Characters should include motivations to be believable.

So this brings us to what, in IN, I consider to be the story-level interpretation of world-level events. In other words, how do encounters combine to form adventures? That is, if we summarize each encounter as a sentence, how does that sentence further or change the story structure?


As a related parallel, I would say the main challenges of IN include:
1) defining the set of possible verbs widely enough that the player doesn't feel constrained yet so that the world can appropriately handle/respond to all (or at least all logical) combinations of verb and world object,
2) providing NPC motivations that lead to believable responses to both PC and other NPC actions, and
3) modeling the story so that the system can direct the story to some conclusion.

I believe point 1 can be overcome largely through brute force--coding up a complex rule-based interaction system. Similarly with point 2, though 2 is harder since NPCs have more complex internal states than objects and must also present their reactions in an audience-interpretable manner. But I find point 3 to be the most challenging. In particular, how can a computer understand what world-level actions mean at a story level, both when interpreting user actions and then when directing NPC/world responses? And, secondly, what model of story is the IN system trying to follow as it directs the action?

In terms of RPGs, human ability fills in a lot of these gaps for us. As mentioned, the rules tell us how to adjudicate the verbs (1). But where they do not, the GM can devise a substitute or house rule. NPC motivation (2) is provided by the GM as necessitated by the story. Story structure (3) is what I'm interested in here. I believe that, for a human GM given a story structure (providing constraints/direction) and a rule system and game world (providing affordances/material), the encounter details largely take care of themselves.


Okay, so in the interest of both IN and RPGs, how do RPG GMs model a story structure? D&Ds DMG suggest two basic approaches: site-based and event-based.

In a site-based adventure, encounters are laid out as a map. Site-based adventures are easy to run since both constraints and affordances are easy to communicate: certain spells aside, players have to move through the passageways, encountering each room in some sort of basic order. Locked doors and other obstacles can further direct them. It's easy to foreshadow what's coming up around the next bend. Choices are clear--do we go left or right at this branch? What do we do with the troll currently picking his nose in the next room?

The DMG points out that site-based adventures can be static or dynamic. Static dungeons don't change--there's a troll picking his nose in this room regardless of what the PCs did in the previous room. Dynamic dungeons change or respond to earlier PC actions. Perhaps the nose-picking troll is summoned from his room to sounds of battle, or perhaps he's warned by a fleeing guard that the PCs let get away. Or maybe the troll simply has a random chance of being asleep rather than nose-picking.

For event-based adventures, the DMG suggests starting with a goal for the PCS, and then consider their opposition. It's important to consider PC motivation and how to entice them into the adventure. Then the GM flowcharts the adventure--what events happen in what order. Again, encounters have an if/then structure, so remember preconditions and stress the consequences of PC actions.

As with site-based adventures, event-based adventures can be static--these things happen if the PCs do this. But they can also be dynamic (called timelines): this will happen at this time, regardless of what the PCs have done so far. Usually you end up with some combo--as in the advancing horde reaches the city on day 3 of the adventure, but the results of that battle will depend on what the PCs did before it got there.


What we can see here is that RPG story planning essentially means creating a directed graph. Each encounter is node. The story moves from one node/encounter to another based on what the PCs decide to do.

In a site-based adventure, the graph will have essentially the same shape as the dungeon--rooms connected by corridors. In event-based, we might end up with more various shapes, such as a linear graph when the GM has a single storyline that the PCs have no ability to change. (These can still be interesting games, as the players have to determine how to advance the story and overcome the obstacles before them. See these rough draft pages of my dissertation for more on potential story forms.)

Of course, there are a couple more concerns than simply the shape of the story graph. The DMG recommends that an adventure should offer a variety of different kinds of encounters (combat, puzzles, social intrigue, etc.) and should make good use of PC abilities. The encounters should be individually exciting, but, together, form the rising and falling action of a story. There should be at least some encounters that offer the players a chance to significantly affect the direction of the story.


So, in summary, we can say that an RPG GM works from a directed graph (often of the branching tree variety) representing all potential stories. Each node is an encounter, and the mechanics of those encounters are determined by the rules system and story world (setting, NPC motivations, etc.). But the actual story produced is a linear traversal of that graph as determined by player choices at each node. This completed linear story should then have appropriate story structure--coherence, rising and falling tension, etc.

So how can we speed the production of this guiding story graph?