Chris Crawford on Interactive Storytelling (67 page)

Values
: Moral dimensions applying to segments; for example, if the segment requires an actor to tell a lie, and the actor has a high Value for honesty, a conflict is created within the actor.

Volition is expressed through nine high-level verbs: decide, inform, incite, dissuade, accept, refuse, perform, condemn, and congratulate. These nine verbs revolve around the concept of making difficult decisions. The first verb, decide, lies at the center of this constellation of verbs; the verb perform represents the actual implementation of the decision. Incite and dissuade are meant to allow another actor to influence another actor’s decision, either positively or negatively. Accept and refuse are reactions to incite and dissuade. Finally, after a decision
is performed, another actor can react to that event with the verbs condemn or congratulate.

Note that this decision structure is boolean at heart, meaning an actor doesn’t choose among various options. Instead, the actor decides whether to take a single course of action. Although this decision is theoretically indistinguishable from the more normal choice among several options, in practice, it just doesn’t match the way characters in drama behave. They often have multiple choices, and they don’t run down the laundry list of choices, considering and rejecting each in turn until they come to the one option they like.

The sample storyworld that Szilas offers has five actors, two goals, three tasks, two obstacles, and two values. The underlying story concerns a murder. A witness, Mr. D, claims that he saw Anna fleeing the murder scene. Joe, the protagonist, is Anna’s husband, and their two friends are Bill and Sylvie. The various components of the storyworld are thus:

Actors: Mr. D, Anna, Joe, Bill, and Sylvie.

Goals: Make witness retract testimony, get rich.

Tasks: Kill the witness, bribe the witness, rob the bank.

Obstacles: Not enough money to bribe witness, police don’t believe retraction.

Values: Nonviolence, respect for the law.

Here is a fragment of a story generated by IDTension and this storyworld. I have translated the broken English of the original into more fluent English:

Anna tells Joe he could try to bribe Mr D, the witness.

Joe accepts Anna’s advice.

Bill advises Joe to kill Mr D.

Joe refuses to do that.

Joe tells Anna he might kill Mr D.

Anna advises him not to do so.

Joe tells Anna he wants to bribe Mr D.

Anna advises him to do so.

Bill advises Joe to kill Mr D.

Joe meets Mr D.

He offers Mr.D some money for changing his testimony, but Mr D. wants a lot of money, and Joe is not rich enough.

He then decides to get money.

He tells Anna about it.

Joe tells Sylvie he might kill Mr D.

Sylvie advises him not to do so.

Joe tells her he is trying to get money.

He tells Bill about it.

Bill advises him that he could rob the bank.

Joe accepts.

He tells Anna he wants to rob the bank.

Anna advises him to do it.

He tells Bill about his intention.

Bill advises him to do it.

He tells Sylvie about his intention.

Sylvie advises him not to do so.

Joe robs the bank, and he is successful.

He pays Mr D.

He obtains Mr D’s withdrawal of testimony.

This story demonstrates some of the features of the IDTension technology. Joe is able to draw inferences about the need to gain money to pay off Mr. D and to
weigh his friends’ advice against his own values. All in all, IDTension works as a logical exercise. Its dramatic content, of course, is low, but we can’t expect much of a first attempt. Interestingly, Szilas reports the same difficulties building an actual storyworld that other authors report (and that I experienced). Apparently, building storyworlds isn’t as easy as building interactive storytelling engines.

Virtual Storyteller
⁸
, a project carried out at the University of Twente in the Netherlands, is part of a much larger project, AVEIRO, which aims to present a complete virtual theater, including storytelling, music, and dancing. The authors built the system as a set of autonomous agents organized by a
director agent
(a drama manager). In other words, each storyworld is populated with a complement of actors who operate with their own goals, but before they can act, they must obtain the approval of the director agent, whose task is to maintain plot integrity. The actors are responsible for the consistency of their own behavior.

The director agent has three means of affecting actors’ behaviors. First is disapproval of their proposed actions; if the director agent disapproves of an actor’s proposal, the actor won’t carry out his or her plan. Approval is based on whether the actor’s proposal is consistent with the plot’s overall direction.

The second means of control is motivational; the director agent can give an actor a goal to pursue. The director agent can’t command an actor to perform a specific action, as that action might not be consistent with the actor’s personality. Therefore, the director agent provides only the goals the actors are meant to attain.

The third means of control is environmental: The director-agent can introduce new actors or objects into the storyworld. These new factors would presumably influence the original actors’ decisions in a direction favorable to the plot.

The authors don’t describe how all this is done. They state: “The director is able to judge whether a character’s intended action fits into the plot structure, using both storyworld knowledge and general knowledge about what makes a ‘good’ plot.” They use a standard rule-based reasoning system, which uses an inference engine that references a set of facts and rules the authors specify. These facts and rules aren’t described, however.

Their system is not designed for interactivity; the director’s role is too intrusive to accommodate the needs of an independent player. Obviously, the director can’t order players to accept a new goal. It might be possible to achieve interactivity by shielding the player from the director agent’s interference. As far as I can tell, no presentable storyworld has been built using this system.

Karin Berndtsson and Lena Kindmark, the authors of InterTale
⁹
, came up with an interesting idea: Sift through a large body of similar stories and find a group of stories with similar architecture. They used Grimm’s fairy tales as their database of stories and isolated about 40 stories they thought had similar structures. They called them “stories with a turning point.” In each story, three people attempted to solve some problem. The first two tried obvious approaches and failed, but the third tried something different and succeeded. Berndtsson and Kindmark built a narrative grammar based on this pattern.