I’d like to step back from the details of the technology and make an important general point about the nature of any technology, not just interactive storytelling technology.
The essence of technology lies in approximation. Awestruck by the microscopically close approximations that some technologies have achieved, we tend to overlook their essentially approximative nature. Yes, machinists can shape a part to within a hundredth of a millimeter; chemists can create a substance whose impurities are measured in parts per billion; chip designers can lay out their chips with resolutions below a micron. But we must remember that these are still approximations that work only because there’s plenty of room for error. The transistors in microprocessors operate on flows of some thousands of electrons when the theoretical limit is a single electron -- what a waste of electrons! There is no such thing as a "pure" chemical solution: every compound has impurities. We get away with the approximation of purity because the impurities do not significantly threaten the task at hand. But "significantly" is just a synonym for "approximately".
The idea behind all of these successful technologies is that the designers have come up with a simplified representation of reality that permits them to apply simple-minded thinking to the task. Let’s think about the ridiculously simple task of precision machining. We want to take this piece of metal and shape it into a cube with accurate dimensions. So we mount it on a milling machine and start cutting it. But what exactly happens when the bit strikes the metal surface? The point of impact gets hot; do the metal atoms vaporize and evaporate away from the metal? Or do we get localized hot spots that spall off of the cooler metal? Perhaps the bit pushes chunks of metal loose rather like a bulldozer. And what if the metal part is not perfectly uniform in internal structure? What happens at the irregularities? Will the metal always flake off at these weakest points?
My point here is that we can, if we choose, view this process in all its glorified complexity, in which case we end up convincing ourselves that it must be technologically impossible to shape the metal part. But in fact, we don’t do that. A machinist doesn’t worry about lattice structures, domains, covalent bonding, color centers, microscopic temperature distributions, and so forth. He uses a simple mental model of the metal, and he shapes that metal cube just fine. The approximation works for him.
The big idea here is the deliberate substitution of a workable approximation for the impossibly complex reality. A machinist doesn’t need no stinkin solid state physics to do his job perfectly well. When tackling more complex subjects, we have to exercise greater care, but the creation of any such technology will surely require a deliberate act of reduction.
Consider the law as an approximation of ethics. We all agree that the law is a poor approximation, but it’s the best we have, and we continually refine it to address increasingly finer points of morality. Consider, for example, our refinement of the ethics of risk. We have, after much experimentation, established a legal principle that imposing risk on others is itself a crime, even if no actual injury is inflicted. An intoxicated person commits a crime just by driving a car, even if nobody is hurt, because the risk he imposes on others is unacceptable. This is an approximation, to be sure. Surely there are some people who can drive safely with a blood alcohol level of 0.10, even though most people can’t. When we convict such a person of a crime, we have imposed an approximation that might not be accurate or even fair -- but we accept the necessity of approximation.
The same thing applies to interactive storytelling technology. When first we look at dramatic technology, our shoulders wilt in the recognition of our towering ignorance. Storytelling is such a complex activity; how could we ever reduce it to a workable approximation? My answer is: with great doggedness. The intellectual cleavage lines of interactive storytelling cannot be discovered by simple inspection; the only way to find them is to roll up our sleeves, get to work, and bung our fingers doing things wrong. In other words, we just plunge in and act like we know what we’re doing, stumbling and slipping at every step, but in the process of failing we learn what really works. And so we go forward. Babies use the technique to learn everything: walking, talking, eating, defecating. Why do we have to be so all-fired omniscient before we attempt anything?
Here’s an example of a cleavage line I recently discovered in interactive storytelling: all dramatic activity consists of a subject doing some action on a direct object. It’s one of those ideas that looks obvious in retrospect, and in fact I recognized its importance at the very outset, but I never realized how universally the principle must be applied. For example, consider the following dramatic sequence: 1. Snidely Adams stabs his friend in the back. 2. A bolt of lighting strikes Snidely Adams. Now, the first sentence readily fits into the subject/verb/dirobject mold. But what about the second sentence? The literal subject is a lightning bolt, but we can’t use that, for a lightning bolt is not a character and has no dramatic volition. We require that the subject and dirobject of every sentence be characters with volition; that’s the only way that dramatic logic can be brought to bear. Clearly, we need to recast the sentence as "Somebody struck Snidely Adams with a lightning bolt." But now the question becomes "Who?" Clearly, Snidely did not strike himself. It’s just as clear that his friend didn’t do it, because his friend is dead.
The solution is to create a character called Fate, who personifies dramatic justice. We then have Fate do the job: "Fate struck Snidely Adams with a lightning bolt."
I concede that this use of Fate is an artifice, and whenever one starts creating artifices, one must question the foundation on which they are built. And in fact I have experimented with a variety of alternatives, thinking them better, and have found -- the hard way -- that such artifices introduce larger problems. The Fate artifice is complete, consistent, and broadly applicable.
How then do we deal with a character’s internal workings? "Jones considered his options: should he lash out or save his response for later?" The literal direct object of this sentence is "options"; once again we have a non-character sitting in the dirobject position. The solution here is to recast the sentence as follows: Jones considered-options-of Jones. In other words, such internal operations take one person as both subject and dirobject. This might seem odd, but it works perfectly well.
Here’s another approximation: the sentence structure we use for relating dramatic activity does not permit many of the more elegant grammatical structures available in natural language. Sorry, there’s no "Behold!"; instead, we say, "Subject SaysBeholdTo DirObject". There are no prepositional phrases; prepositions must be built into the verb, and their objects stored in additional terms I call "secondary objects". There are a dozen such secondary objects for every sentence: two of each for each of the fundamental data types of the technology (characters, verbs, events, things, numbers, and logicals). Most of the time we skip over those secondary objects and leave them empty, but they are occasionally necessary. For example, in the sentence "Subject Gives DirObject ThisThing", that last term shows up in a slot called ThingObject1.
What I have done here is to reduce the richness of natural language to an approximation that works well enough for dramatic purposes. It’s not eloquent, but it works.
At this point we encounter the classic objection to all technology: it’s an approximation and does not capture the richness of the reality. I have little patience with this objection; while it is both valid and correct, it is also pointless. All technologies reduce reality to an approximation; the fact of approximation is not a cause for objection -- only the degree of approximation. I concede that we will never capture all the nuance of drama and language in such an approximation, but I know from direct experience that such a system can be used to build an interactive storyworld with dramatically interesting behavior. A chess-playing program on a computer will never understand chess, nor will it capture the elegance and style of a great chess master -- but it can still beat most players. The fact that interactive storytelling cannot rival the work of Shakespeare is no cause for despair. Surely there is some middle ground between the Bard and Nintendo.