Chris Crawford on Interactive Storytelling (14 page)

So a simple version of the Arthurian storyworld includes the love triangle, the search for the Grail, and the final battle with Mordred. At the outset of the storyworld, young Mordred is seeking to unseat Arthur as king; Lancelot has just
arrived at Camelot; Guinevere has been Arthur’s queen for some years; and the Saxons are continually raiding the lands of Arthur’s people, killing peasants and stealing cattle. Arthur must organize war parties, lead counterattacks against the Saxons, fight battles against them, and drive them back. He must fairly distribute the spoils of war among his subordinates, rewarding those who fight well without antagonizing others. Mordred attempts to foment trouble by playing on the rivalries and resentments of those who believe Arthur hasn’t properly recognized them. Arthur must also adjudicate disputes among his people, which sometimes requires more diplomacy than justice. Meanwhile, Guinevere, as queen, is responsible for running Camelot, and her activities generate new difficulties demanding Arthur’s interventions. If Guinevere feels unsupported by Arthur, her resentment might flower into a relationship with Lancelot—assuming that he too is somewhat disenchanted with Arthur. Or perhaps it won’t. Either way, Mordred spreads tales besmirching Arthur’s reputation and Guinevere’s purity. The situation comes to a head when Mordred declares a rebellion against Arthur. The nobles choose sides, based largely on their respect and affection for Arthur. In a final battle, numbers will likely determine the outcome. Of course, if Arthur has maintained peace and prosperity for all, Mordred will never get his chance and Arthur will die in his bed, beloved of all.

This example shows only one way to build a storyworld. There are many other possibilities. The key realization is that a storyworld is a much larger creation than a story. This makes perfect sense when you realize a simple theorem about storyworlds: A single playing of any storyworld generates a single story. In other words, when a player goes through a storyworld, he produces a story. The player’s path is a linear sequence of events—a
storyline
. He could relate his experiences to another person, and the listener would perceive the player’s experience as a story. Different playings of the storyworld can yield many different stories, however. Hence, a storyworld contains zillions of incipient stories. Clearly, this creation is much larger than any single story. Building storyworlds is necessarily a huge task.

No matter what technology is used for interactive storytelling, it must somehow assemble pieces into a complete story. There’s one exception to this simple statement, however. It’s possible to think of a computer story technology in which players make all their decisions at the outset of the story. Perhaps a player specifies some character traits for the protagonist; perhaps she answers a questionnaire. Whatever method is used, the player’s contribution is complete before the story begins. The software then creates a story using these specified parameters. This technology would certainly be interesting, but I wouldn’t accept it as interactive storytelling because it’s not interactive. The player speaks just once to the computer, and then the computer thinks it over and speaks back to the player with a long story. Interaction requires an alternating cycle of speaking, listening, and thinking. One step doesn’t cut the mustard.

For true interactive storytelling, however, you can be sure that the computer assembles the complete story out of little story fragments. This process is fundamental to computing. A computer sees an image as a big collection of little image fragments called
pixels
and manipulates the image by changing the individual pixels. A word processor sees a document as a collection of text characters mixed with a set of formatting commands. Deep inside the heart of every computer program is at least one central quantity or concept broken up into lots of tiny fragments, and the computer executes processes on this set of fragments to generate its output. This concept is absolutely fundamental to all computing; it applies to interactive storytelling just as well. The question, then, is “What are those little story fragments that the computer plays with?” In other words, what are the atoms of storytelling?

I use the term
substory
to describe the atoms of storytelling, but giving these atoms a name doesn’t answer the question. What are substories?

A substory is a single dramatic step; it’s an event or a change. It can be described in a sentence that specifies an event: “Jan realized the significance of the white dog” or “Andre lived happily in his new life.” This event can be tiny (“James sidestepped the descending sword and swung to his right”) or big (“James killed Thomas in a swordfight”).

Note that substories aren’t measured with a clock. That is, a substory cannot be a chunk of time. Stories don’t move forward in synch with the clock; they crawl through moments of intense emotion, and then leap forward hours, days, even years. Some substories can take place in a flash (“In an instant, Jan realized the significance of the white dog”); others can span years (“and so Andre lived happily in his new life, until one day…”).

These examples show that the data structure defining a substory is the same as that for a sentence. There must be a subject, a verb, and, most likely, a direct object. A substory probably also includes additional information in the form of prepositional phrases, adverbial phrases, and so forth. Its basic form, however, must be functionally similar to a sentence.

What about exposition? Every story has to provide context and background through straightforward exposition, and exposition doesn’t necessarily involve events. For example, “The cavern was astounding; on the walls, fist-sized jewels glistered in the light of their torches, while the path on which they walked was paved with golden bricks.”

This substory doesn’t fit the standard sentence template you described. It doesn’t represent any event or change. It’s a state. Doesn’t this invalidate your description of a substory?

I don’t think so. While working on the Erasmatron, I did encounter some problems with exposition because exposition must never be repeated. So I designed a special object called an
exposition
to describe some stage, actor, or prop when first encountered. But expositions never interfere with the flow of the story; the story development is independent of expositions. An exposition is, in effect, an unrepeatable substory.

How big should substories be? At what level of dramatic resolution should players experience the storyworld? At one extreme, you could offer players the opportunity to make only the key decisions in the storyworld, such as the six decisions that Luke Skywalker made. This method would keep the storyworld small, but would be unsatisfying to players, as the storyworld wouldn’t offer many variations on the basic storyline. In the Luke Skywalker example, all six
decisions are simple binary decisions; therefore, the story has (theoretically speaking) only 64 possible outcomes, and in practice players would experience far fewer. After all, should a player decide not to take Obi-Wan partway to Mos Eisely, he returns home and is killed along with Uncle Owen and Aunt Berue. End of story. In truth, the original story has only eight variations, one of which matches the original movie. The other seven all end abruptly with the player being killed or living out the remainder of his life in obscurity. Therefore, restricting a storyworld to only gigantic decisions keeps things too simple to be interesting. Perhaps a simple system would be useful for children first experiencing interactive storytelling, but adults would find it dull.

At the other extreme is something like a stream of consciousness. Every experience, thought, and perception is treated as a substory and presented to the player. As with conventional stories using stream of consciousness, the result can be overwhelming, impossible to follow, and tedious.

My belief is that the size of substories is one of the fundamental artistic decisions every storybuilder must make. Dramatic emphasis is achieved by using small substories; tedious material can be compressed into a single larger substory. The storybuilder’s ambition and effort determine the total number of substories in any storyworld. Tolstoy could have written
War and Peace
as a short story or regular-size novel, but he chose to write a behemoth. To each his own.

You could refer to traditional story forms to determine the ideal scale of dramatic resolution. In literature, you can choose between a sentence and a paragraph as your atoms. A typical novel has so many sentences, however, that using paragraphs is preferable; the number of paragraphs in a novel seems a manageable and appropriate number of decisions for the player to make. Again, you see that the length of a substory is variable; some paragraphs in a novel are short and some are long. Comics provide a particularly useful angle on the problem of determining the ideal scale. You can view a single frame of a comic as a substory; no other medium has such a cleanly quantifiable structure.

The conclusion is that the ideal resolution for interactive storytelling is somewhere between literature’s sentence and paragraph. However, it’s too early to close this topic; I suspect that creative artists will expand the boundaries of this concept.

Some elements of traditional storytelling cannot be used in interactive storytelling. For example, you can’t have what I call the “Third Option”: the protagonist’s stroke of genius that gives him a third option for escaping the sharp horns of a dilemma.

Here’s a somewhat obscure example requiring some explanation. In the movie
Bless the Child
, a devil-worshiping cultist accosts a God-fearing little girl on the roof of a high building. He taunts her faith in God and forces a test upon her. Placing her on the ledge of the roof, he demands that she prove her faith. She must leap off the roof and trust God to save her, or she must reject God and declare her faith in the Devil by taking the hand of the cultist. A dark shroud of menace hangs over the scene; neither the audience nor the little girl can guess whether the cultist will throw her off the roof if she refuses to make the impossible choice he is forcing upon her. After a nerve-wracking pause, the little girl turns to the cultist, smiles, and says invitingly “After you!”

This scene actually contains two important lessons about interactive storytelling. First, consider the long logical train her response triggers. She is suggesting that the cultist should prove his own faith in his own god (the Devil) by making exactly the same leap he’s trying to force on her. If the cultist is unwilling to leap off the building, how can he ask the girl to do so? He is revealed as a hypocrite, and his either-or choice is ruined. It’s a complex logical twist that comes as a complete surprise to the audience. Although it’s impressive work, I think you can discount the possibility of a computer coming up with a similar option. It’s simply too hairy and requires too deep an understanding of human nature. Perhaps next century.

Second, even if an interactive storytelling system could concoct such a stroke of genius, it couldn’t offer it to the player. If the system were to present the girl’s choices in that form, the list would read:

Leap off the building to your death.

Reject God and embrace the Devil.