Book Review: Chris Crawford on Interactive Storytelling

Interactive Storytelling

Deep in an ancient forest, you slide the strange metallic boxes across the moss-covered earth. You push the blue cube into the corner of the glowing rectangle branded into the ground… say the magic words… and hear a loud crack as the power grid clicks into action! The old wizard's projection suddenly appears and begs you to rescue him – revealing, at long last, that HE is your long-last grandfather. But if you go after him, you know the Sacred Chalice will fall into enemy hands and be lost forever. The choice… is yours.

Is this Interactive Storytelling? Most game developers would probably say so. But according to Chris Crawford, this kind of storytelling is anything but interactive.

Crawford, a game design pioneer turned interactive storytelling researcher, uses the term “Interactive Storytelling” to refer to a type of entertainment experience that is much more than a video game with narrative thrown on top. True interactivity, according to Crawford, means a story that is literally created out of the decisions made by a player, moment to moment, scene to scene.

The challenges and solutions that lay behind the quest for such a story engine is the subject of Crawford's new book, Chris Crawford on Interactive Storytelling.

In the book, Crawford sets forth his vision for a world in which interactive stories could become a dominant form of mainstream entertainment. Examining in detail the many challenges and rewards implicit in the development of an interactive story engine, Crawford takes us through the past, present, and future of this emerging medium, and guides us through his set of solutions for implementing the interactive storytelling engine.

Abstracting the Storytelling Process

So what is an interactive story? According to Crawford, it begins with understanding the concept of a storyworld.

Story, says Crawford, is the presentation of a sequence of events that form a recognizable pattern to the human mind, one that summons our deeply ingrained set of cultural archetypes and teaches lessons in the telling. But interactivity is almost antithetical to this goal, because allowing a user to make choices that affect a story is almost guaranteed to upset the careful balance that goes into crafting that carefully structured pattern.

In Interactive Storytelling, the story engine must be able to generate stories on the fly, based on user feedback, without relying on pre-scripted action or a hardwired series of events. Pulling this off requires a deep abstraction of the storytelling process, and a storyworld is the space in which that abstraction can occur.

In a storyworld, a player is able to make choices that determine the future course of that world's events. The choices must be dramatically interesting, varied, and balanced – meaning that the story actually can develop in many different directions based on the player's decisions. Yet through it all, says Crawford, the storytelling engine should be capable of maintaining the crucial sense of structural integrity that turns a simple series of events into a tried and true story.

The Erasmatron, Crawford's own interactive storytelling system, is addressed in the book, but Crawford presents his implementation strategies in a more abstracted, generalized way. By discussing the theory and common sense behind his design concepts, Crawford paint for us a picture of how an Interactive Storytelling engine might look, and what the obstacles might be that a designer would face in creating a realistic and satisfying storyworld.

Story and Natural Language

Natural language – simple English sentences – form the core of Crawford's storytelling system. In his own lingo, a single story element (or beat) is called an Event – a moment in which a Subject performs a Verb upon an Object . In its simplest form, this could be as basic as “Jane slaps Bill.”

The Subject in that sentence is what Crawford calls an Actor – usually an in-game character. (But not always – “Fate” can also be an Actor, allowing for coincidence and catastrophe in any given story.) Actors come with all types of interesting numerical and verbal data attached to them, describing their emotional proclivities, behavioral tendencies, and relationships to other characters – as well as what “roles” they're playing at any given time, and the “gossip” that they're carrying around in their virtual heads.

Verbs , on the other hand, are the actions that an Actor can perform. This is important – because by definition, Verbs are what a player does, in any interactive medium . For a player to always have choices available that are truly dramatic and interesting , a great abundance of verbs would need to be wired into the engine – by Crawford's estimation, thousands of them.

And an Object is any aspect of the world that an Actor can be act upon – such as props, environmental features, and other Actors. Crawford allows for both direct and indirect objects in his grammar – e.g. “Jasmine feeds Gary a cake” – thus allowing for Events in which multiple characters or objects are involved.

Intelligent Decision-Making

Now, this may sound simple enough, but growing this core structure into a workable storytelling system is a formidable task., Crawford has spent a good deal of time thinking of the myriad dilemmas raised by a language-based storytelling engine, as well as several interesting solutions.

Verbs, for instance, are usually contextual. Before “Jane slaps Bill” can take place, an engine would need must check Jane's personality matrix and see if she's really in the right mood to do any slapping. (If Jane is a gentle but frustrated woman, she may simply stomp her feet instead.) Bill's relationship to Jane also factors into the equation – the slap is probably more likely if they're a married couple than if Jane is Bill's grandmother.

Additionally, the engine must know the current state of affairs in an the immediate environment – for instance, whether Jane and Bill are really in the same room together. And the engine must be aware of the history of the storyworld – contained in what Crawford terms a HistoryBook – since Jane slapping Bill would probably occur in response to an earlier stimulus, such as Bill insulting Jane or sleeping around behind her back.

Information is another element that can affect the decisions made by computer-controlled Actors. For instance, Gossip is used to pass information (and mis-information) between in-game characters – so if Gary hears some juicy gossip about Jasmine, he might decide to spill the beans, depending on his moral fiber. But anticipating Jasmine's negative reaction could factor into Gary's decision too – as could the chances that the gossip will damage his reputation once it hits the Grapevine.

In this manner, the seemingly simple sentences of the core engine grammar begin to organize into more cohesive story elements. By no means does Crawford present a bullet-proof system for handling the enormous number of potential events in a complex storyworld – and in many cases, Crawford notes challenges without necessarily presenting solutions – but the ideas he does present certainly make a good start.

Drama Management

Integrating Events into actual stories takes an additional level of data management – via what Crawford calls a Drama Manager .

A Drama Manager is a system that can monitor the progress of a story, determine how the story might develop over time, and implement that determination within the storyworld itself. In short, a Drama Manager is responsible for providing dramatic structure to the events that unfold in a story.

This is a critical function, because storytelling requires a fairly specific structure in order to be understood as a story by an audience. Although Crawford's Drama Management concept is a good idea in theory, it remains to be seen exactly how such a system would organize algorithm-generated Events into a satisfying dramatic structure. It's one thing to create hard-wired networks that determine the relationship between every verb in a virtual dictionary – but another thing entirely to teach a computer to recognize the dramatic value in a dynamically generated event.

This is one area what will need deeper exploration before Crawford's vision for Interactive Storytelling can really become a reality for developers.

A Matter of Theory

It's worth mentioning that Crawford's system is, to date, entirely theoretical. He has developed some of his ideas in the form of the Erasmatron, his own interactive storytelling system, but there has yet to be a product that makes use of his design concepts.

That by no means invalidates his ideas –for Crawford is the first to admit that his system is as yet incomplete and unable to address some very critical design problems. But one does wonder just how feasible an implementation would be at this early stage of the game.

For instance, in spite of Crawford's assertion that interactive storytelling is all about what a user does , it's not clear how a user would actually interface with an interactive story. Selecting verbs from text lists seems like one solution – but that would ultimately reduce the user's choice of Verbs to one: “pick.”

So there's definitely room for Crawford to present some more integrated and developed design concepts in the next go-around.

All told, though, these holes are to be expected, given the relative newness of Crawford's approach to computer-based storytelling. As Crawford himself says, the first adventure games were remarkably simple text-based programs, but led to the game industry as it exists today. In a similar fashion, Crawford hopes that early attempts at Interactive Storytelling will eventually lead to a revolutionary new entertainment genre.

Vision for an Interactive Tomorrow

Beyond the various implementation discussions, Crawford spends a good deal of the book discussing the meaning of interactivity, the past and future of the quest for Interactive Story, the creation of development environments for would-be interactive storytellers, and ongoing research into the potential of interactive story systems.

The book's style, by the way, is quite candid, and Crawford's opinionated personality does make its way into the material. At times, this can be a bit distracting, but his frank style adds color and character to a book that could otherwise have been bogged down by its abundance of theory and partly developed implementation strategies.

Overall, though, “Chris Crawford on Interactive Storytelling” is a comprehensive treatment of a topic that continues to be of interest to game designers and storytellers alike. His ideas need further exploration and refinement, but the book itself is an informative and welcome addition to the wider discussion of the industry's future.

If an Interactive Storytelling revolution is indeed just a few years around the corner, then you can bet Chris Crawford will be at the crest of the wave.

______________________________________________________