Chris Crawford on Interactive Storytelling (42 page)

Laurel’s laundry list is also overlong. Although all these functions are desirable, just three fundamental steps are required for a drama manager, all of which spring from my basic definition of interactivity in
Chapter 2
:

A cyclic process between two or more active agents in which each agent alternately listens, thinks, and speaks.

The drama manager must monitor the story’s progress. This step is simple listening—but listening is not the same thing as hearing. To hear another person speak, you must be able to interpret the noises in terms of the language used. All those squeaks and grunts must be translated into words and sentences, which must then be analyzed to determine their meaning. In the same fashion, a drama manager must interpret the raw material of the developing story (its events) in terms of an abstract drama language. It must recognize the patterns of behavior that differentiate drama from tedium.

The simplest means of accomplishing this step relies on
overview variables
, which are numbers the storytelling engine calculates to assess the storyworld’s overall state. For example, you could calculate the sum of all affections held by all Actors for each other; this number would indicate the overall level of “good feeling” in the storyworld system. A low value means that everybody hates each other; a high value indicates universal love and brotherhood. The drama manager monitors this value to determine what to do. If everybody is too happy, maybe it’s time to toss something ugly into the mix. If the storyworld has degenerated into a dog-eat-dog jungle, perhaps it’s time to bring some balmy sunshine to the cast.

Creating overview variables is a central creative challenge in interactive storytelling. The database for a storyworld will be large: perhaps a megabyte of numeric information. All those numbers can be put together in a staggering number of ways. Here are just a few possibilities for overview variables:

Gross trust:
The sum of all the trust that all Actors feel for each other; at low values, this variable signals a complete breakdown of social cooperation.

Acceptability of violence:
A running sum of violent actions; as this sum grows, Actors’ inhibitions against violent behavior lower.

Dullness:
The sum of verb
Import
values (see
Chapter 13
, “Verbs and Events”) for recent events. If the sum becomes too small, the cast has settled down into a dull period and the storyworld needs some spicing up. Inject some sensational or divisive event.

Storyworld exploration:
Have all the verbs in the storyworld been used at least once? Has the player spent most of the time using a small set of verbs?

Having developed a coherent dramatic interpretation of the events so far, a drama manager must then determine how the story should progress from that point forward. A number of schemes have been tried for drama manager computations.

A well-designed set of overview variables makes this task easier, as each overview variable indicates some significant factor in story development. The overview variables listed previously suggest some action the drama manager should take. Conceivably, a good drama manager would consist of nothing more than a large set of overview variables, each controlling some action by the drama manager. For example, if the Storyworld Exploration overview variable indicates that the player hasn’t used many of the verbs in the storyworld, the drama manager could search through the unused verbs, looking for those verbs that tend to precede the unused verbs. Having identified these
leader verbs
, the drama manager could then generate events using them; this might get the player to travel down some previously unused paths.

Another approach uses
dramatic templates
. The storybuilder generates a large set of templates, each of which constitutes a well-formed story. Then the drama manager matches the story generated so far with each template, searching for the one that best fits the story so far. That template then becomes the guide for action by the drama manager.

A third approach assumes the existence of a
story grammar
, which is a set of rules governing the sequencing of events in a story; it’s analogous to grammar in linguistics, but it concerns the connections between narrative components in a story instead of the connections between words in a sentence. This story grammar then provides the basis for calculating story development. Although much has been written about story grammars, precious little has actually been accomplished. Vladimir Propp’s work described in
Chapter 9
, “Data-Driven Strategies,” is often cited as an example of a story grammar.

A drama manager’s final task is translating its determinations into some form that will actually change the storyworld in a manner that helps the story evolve in the desired direction. You can use any or all of the following schemes to influence story evolution.

The most overbearing technique is making environmental alterations in the storyworld. If a player attempts to go over the mountain pass, and your drama manager determines that she should instead be traversing the tunnels underneath the mountains, the drama manager need only throw a few snowstorms and avalanches at the player. If that doesn’t work, have the trail end abruptly at the top of a cliff. All manner of physical constraints can be applied to force players into the intended course of action. These devices, however, are often transparent and insulting to players. Use them only as a last resort to save the developing story from catastrophe.

In a different set of circumstances, using environmental manipulation is more satisfying:

It’s the climax of the movie. The hero has driven the villain out of his stronghold and pursued him up the craggy mountaintop. The trapped antagonist turns and fights, and the two struggle as the wind tears at their capes and the rain lashes at their faces. A rock gives way under the hero’s feet and he slips, nearly plummeting into the yawning chasm below, but somehow his fingers find a handhold and he dangles helplessly at the feet of the delighted villain. The villain sneers and picks up the hero’s own sword, raising it high over his head to deliver the deathblow. Suddenly, a bolt of lightning strikes the villain, who plunges off the mountain peak, his final scream echoing off the cliffs. The hero picks himself up and gazes down at the still-falling villain.

The difference between the two examples (snowstorm versus lightning bolt) lies in the matter of free will. In the former case, environmental manipulation is used to foil the player’s intentions. The storybuilder has assumed an antagonistic stance with respect to the player. This will never do; storyworlds aren’t tests of will between storybuilders and players. In the latter case, the environmental intervention is an integral part of the narrative, not a blatant attempt to block the player. Indeed, there are actually two environmental interventions in this example: the rock giving way beneath the player’s feet and the lightning bolt striking the villain. This could be interpreted as a one-two punch delivered by a drama manager who has already determined the story’s outcome. The player, having driven the villain out of his stronghold, has successfully concluded the story, and now the only remaining task is to finish off the villain. The drama manager wrings one last bit of drama out of the situation before concluding the story.

A more subtle scheme is instilling some new goal into an Actor. This new goal, however, must of course be compatible with the Actor’s personality; it simply won’t do to have Gandalf suddenly decide he wants to make whoopee with Galadriel. This new goal should also be of a temporary nature so that its effects are localized to the immediate situation. If your player as Frodo is just about to do something seriously wrong, make Samwise cross his legs desperately, rock back and forth, and insist that he absolutely
must
go urinate. Then get Samwise into serious trouble, diverting Frodo from his intended course.

The most elegant solution is altering the personalities of other Actors in such a way as to induce them to make decisions that influence players in the desired direction. This scheme is the most indirect; the drama manager makes changes that will cause events that will change the player. That’s what makes it so elegant—but it also makes it difficult. To continue with the
Lord of the Rings
examples, altering Samwise’s personality can indirectly influence the behavior of the player as Frodo. If Frodo persists in an obviously futile course, increase Samwise’s exasperation with Frodo until Samwise explodes in anger. If Frodo seems unwilling to commit himself to action, increment Samwise’s pluck and optimism so that Samwise will encourage Frodo.