Chris Crawford on Interactive Storytelling (30 page)

Perhaps if you combined all the previous strategies, some synergy among them might contribute to an overall solution. If this thought has crossed your mind, stamp it out! It’s hopeful thinking taken beyond reason. There’s no reason to believe that some magical synergy exists. You can expect synergy only among elements undergoing some mutual interaction. None of the previous techniques shows any indication of interesting mutual interactions. Throwing more and smellier ingredients into the pot doesn’t get you any closer to cooking up gold.

There remains the important and exceptional case of The Sims, Will Wright’s brilliant exercise in simulating household life. Much has been made of this game’s storytelling aspect. Of course, The Sims could be described as an environmental approach to storytelling, although it offers no explicit storytelling, and Will Wright is quick to point out that it’s definitely not an interactive storytelling product. Yet millions of people who play The Sims laud its storytelling content. They perceive story to exist in their characters’ behaviors, and they concoct elaborate stories to explain the activities going on in the game. This creates a fascinating theoretical conundrum: There’s no story in the product, yet players imagine stories based on its behavior. Are these stories real? Can a product that stimulates the player’s imagination to create a story be considered interactive storytelling?

I tend to dismiss such definitional arguments and concentrate on the delivered reality. If customers perceive a story, that’s all that matters—it’s good enough for me. Yet this answer leaves me unsatisfied. People concoct stories in all manner of situations, for all manner of reasons. Children playing with toys create intricate stories around those toys, but toys aren’t storytellers. The enjoyment of players who festoon their play experiences with grand stories of their characters’ life histories can’t be questioned. The Sims, however, doesn’t offer storytelling per se. It is, to use Will’s own description, a toy. Like any good toy, it stimulates the imagination, and that imagination proceeds to create stories.

If you look more closely at The Sims, you can see that it falls far short of anything that could be called interactive storytelling. The characters in The Sims have no personality to speak of, and interactions with other characters are shallow and devoid of emotional content. If you apply Crawford’s First Rule of Software Design, and ask “What does the user DO?” the answer is even more revealing. Players of The Sims guide their characters in going to the bathroom, taking showers, preparing and eating meals, cleaning the dishes, taking out the garbage, cleaning house, sleeping, and earning a living. This is not drama; this is a housekeeping simulation. Alfred Hitchcock once described drama as “life with the dull bits cut out.” The Sims is life with the dramatic bits removed.

I do not offer these cruel observations to denigrate Will Wright’s achievement; anybody who throws mud at such a shining triumph besmirches only himself. My intent is to make clear that, for all its greatness, The Sims isn’t interactive storytelling. It’s a brilliant success—but not an interactive storytelling success.

Might The Sims lead game designers in the direction of interactive storytelling? Doesn’t it offer a solid foundation on which to work?

I don’t think so. Six years have elapsed since The Sims first appeared. During that time, Electronic Arts has released a cavalcade of expansion packs pushing the design in many different directions: Livin’ Large, House Party, Hot Date, Vacation, Unleashed, Superstar, and Makin’ Magic. Yet none of these expansion packs expands the design’s dramatic range; they merely add more parts to the simulation. If after six years no progress has been made toward better dramatic interaction, there’s little hope we’ll see any in the next six years—or the six years after that.

The environmental approach is a dead end; it doesn’t solve the problem of interactive storytelling. You can set up a gigantic stage, equip it with a cavalcade of fascinating props, create spectacular scenery and magnificent sound effects, and still have nothing. Actors make the stage come alive, actors with talent and perception. A good acting company can make Shakespeare come alive on a plain wooden stage; an army of cardboard figures in a magnificent 3D environment is still just a pile of cardboard.

ONE OF THE TIME-HONORED
strategies in algorithm design is the use of
data-driven
methods. Sometimes coming up with equations that carry out the required calculations is just too difficult; in some of these cases, using data-driven methods is possible. These methods substitute tables of data for formulae.

A data-driven storytelling engine would require two major sections: a mass of data and a means of assembling that data into a story in response to the player’s actions. The data itself is composed of two parts: story components (the basic parts of the story) and connectivity data (how these parts connect with each other).

Story components
are bits and pieces of stories that the engine stitches together to yield a continuous story in response to the player’s input. They can take a variety of forms. The most basic forms are simple nouns: Actors, Props, and Stages. They would be characterized by their dramatic functions, which in turn might form the basis of the connectivity data. For example, an Actor could be characterized not in terms of personality variables, as discussed in
Chapter 11
, “Personality Models,” but in terms of actions. Thus, Snidely Whiplash might be characterized in terms of willingness to do dastardly deeds. If the story calls for a dastardly deed, the engine searches through the Actors and finds Snidely Whiplash ready to perform.

In other words, the connectivity data for noun-based story components would specify the types of verbs those story components require. This in turn requires a classification system for verbs based on noun requirements. For example, verbs might be classified by their “Dastardliness,” their “Nurturative Value,” their “Humor,” and so forth. Then a story component would call for certain combinations of these dimensions. I haven’t explored the possibilities of such a system.

Another story component could be more detailed dramatic archetypes (or clichés, if you prefer). Actor instances of these archetypes could be Rebel With a Chip on His Shoulder, Female Detective’s Client With Great Legs, Loner With a Dark Past, and so forth. Prop instances might include Fabulously Valuable Archaeological Artifact, Tattered Old Treasure Map, Knife in Boot, or Handy Getaway Vehicle. And for Stages, you have settings such as Dark Threatening Forest, Smoke-Filled Bar, Homey Dining Room, and Shabby Office.

I’m more optimistic about story components based on events: The First Kiss, Escape on Horseback, Hide From Pursuer, Accidentally Reveal Secret, and so forth. Indeed, my Erasmatron technology relies on this concept. However, creating the connectivity data for these story components is a time-consuming task.

Connectivity data
provides the information the engine uses to decide which story components to stitch together. This data is the yin to story component’s yang, so it is difficult to discuss connectivity data in the absence of a specific set of story components. Moreover, connectivity data is more difficult to imagine than story
components because it’s necessarily more abstract in nature. Connectivity data concerns cause-and-effect relationships, not simple existences. All the ideas I offered for story components are easily recognizable, but the connectivity data for those components is harder to explain. It’s simple enough to talk about evil witches, gingerbread houses, or ovens, but declaring the connections among these components takes a lot more work.

This brings me to a digression: In any project, tackling the most difficult task first is always prudent. Many designers put off the most difficult task, hoping that if they can start off by solving the simpler problems, they will create momentum or credibility for the project that will enable them to bull through the tough parts. This is a mistake, for two reasons.

First, if you can’t solve the tough problem, the whole project founders and the time you have spent on the easy problems has been wasted. Better to kill the project early than magnify the failure by putting off the day of reckoning.

Second, every design decision you make constrains all future design decisions. If you’re designing a superfast inkjet printer, and you design the case first, the dimensions imposed by that case control the design of the printing mechanism, which could seriously hamper your design efforts. If, however, you design the printing mechanism first, you can easily design a case to fit over that mechanism.

But automotive designers often design the car’s exterior first, and then design the engine to fit into the engine compartment specified by the exterior design
.

That’s because designing automotive engines isn’t as difficult a task as designing appealing exteriors. Automotive engineers have been laying out engines inside engine compartments for a century now; the design trade-offs involved are well understood. Appealing exteriors, by contrast, are considered to be a killer problem in automotive design. The sequence isn’t inside to outside; it’s more difficult to less difficult.

Therefore, the connectivity system should be designed first. What elements make a good connectivity system? I can’t offer any clean answers to this question; I can only present some existing sources in the following sections that might be worthy of further exploration.

Folktales, or “fairy tales,” as they are commonly referred to, share many common themes, elements, and plots. Carl Jung studied folktales carefully to determine what they revealed about the human psyche; later, Joseph Campbell carried on Jung’s work in a less rarefied style. Campbell showed that common psychological threads do indeed run through folktales the world over.

Similarly, folktales seem to share many common components, or
motifs
, as scholars call them. Evil witches, nasty spells or potions, young innocents, journeys to faraway places—these and other motifs seem to crop up over and over.

In trying to understand all these connections, scholars have attempted to organize the vast database of information about folktales in some fashion that reveals their underlying commonalities. A hundred years ago, a Finnish scholar named Antti Aarne prepared an index of the various types of folktales and their motifs. Later in the twentieth century, an American researcher named Stith Thompson greatly expanded on the index Aarne had created. The result was a sort of Dewey Decimal catalogue of folktales, which a number of scholars have subsequently built on. The resulting index goes by the name of
The Aarne-Thompson Types of the Folktale
. Here is the highest level of the index: