| || |
The City of IF Story
The City of IF is a web site dedicated to the art of interactive storytelling, in which author and audience create a story together. The site was created by Mark Keavney, PhD, a web designer, writer, and longtime roleplaying game player. Below is a six-part history of the City of IF, which includes some general thoughts on myth, roleplaying, and interactive storytelling, and a vision for the future.
To contact the City of IF, email firstname.lastname@example.org.
Lost in the Wilderness: The Many Paths of Interactive Storytelling
1. Myth and Roleplaying Games: Seven Treasures and Five Dragons
2. Lost in the Wilderness: The Many Paths of Interactive Storytelling
3. Storygaming: On This Land I Will Build a City
4. The City of IF: Past and Present
5. The City of the Future
In the last section I talked about my background as a roleplaying game player—what I loved about those games and why I eventually stopped playing them and instead started to look for a way to bring their "treasures" to a broader audience. In this section I'll describe what I found about other attempts to do similar things—and why in the end I chose a different approach from other free RPG games.
When I started to look around at the interactivity and storytelling, I found a crowd of people interested in this. They came from many places - from tabletop roleplaying, free multiplayer online RPGs, computer games, drama, art, computer science - and they called what they were searching for by many names—interactive storytelling, cyberdrama, digital media, interactive fiction—but their interests had a strong family resemblance, and it was the same family that I'd known in roleplaying. Probably the best way to describe our common goal was that we were trying to create a new medium that combined the narrative of stories with the interactivity of games.
So I read papers, and exchanged emails, and played with lots of demos and prototypes. I traveled to an interactive storytelling conference in Europe (TIDSE 2004—Technologies for Interactive Digital Storytelling and Entertainment) to hear the latest thinking first-hand. If you're interested in exploring the same territory yourself, a few good places to start are Chris Crawford's extensive library of writings on the subject, Nick Yee's site on the psychology of MMORPGs, the site for a project called Facade (which I'll talk more about later), the interactive fiction magazine XYZZY News, and Grand Text Auto, a group blog about "machine narrative, games, poetry, and art."
I don't have the space for a complete record of everything I learned. But I'll describe three different general approaches to the challenge of interactive storytelling, to give you a sense of what's out there in the wilderness.
The most straightforward way to make a story interactive, and the one that most people think of first, is to create multiple branching versions of a story and let the players choose one branch or another as they go through the story. So an interactive book, for example, would always start the same way, but the readers would be directed to different pages for the middle and ending of the story depending on what they chose.
This idea generated a lot of excitement in the 80s and 90s. It was the basis of a book series called Choose-your-own-adventure books, it was used by some video games (e.g. Wing Commander) to add narrative to the gameplay, and it was the inspiration for many current U.S. patents on different ways to incorporate branching storylines into movies and TV.
But this approach has not really gone anywhere, and I'll explain why (my analysis owes much to Chris Crawford's excellent and much more detailed article on the limits of branching storylines). The fundamental problem with branching storylines is that the size of the story (and therefore the resources required to create it) increases exponentially with the number of choices. For example, imagine a simple short story with only three places for the reader to choose, and only three different choices at each one. Such a limited amount of choice would probably not feel very interactive to the reader, but even this small amount of interactivity would require the author to write (or draw, or film) 3x3x3 = 27 different endings. A story with six decision points of four choices each (still a modest amount of interactivity, far less than what might occur in a novel-length interactive story) would require 4x4x4x4x4x4 = 4096 different endings. This is clearly way beyond what is practical for an author to create in any medium.
There are several ways to try to get around this problem. The most common is to "fold back" the storylines into each other, so that two different branches would lead back to the same place. For example, if you meet a man-eating giant you might choose to trick it, fight it, or hide until it falls asleep, but no matter what you do you'd get past the giant and end up at the same place. Or maybe there are six choices: in three you get past the giant while in three it chases you away. All of the applications I mentioned above used some degree of foldback to keep the number of storylines manageable.
The main problem with fold-back is a subtle one: it imposes a constraint upon the story author that tends to drain the story of its emotional life. The constraint is that the author has to write each piece of the story without knowing what came before. When creating the scene after the man-eating giant, for example, the author might know that the character got past a giant, but not how. Without the details, the author can't refer to that incident in any emotionally meaningful way: can't build upon the character's courage in fighting the giant, or have the character overcome the fear that caused him to hide from it. Stories written in this way may make sense intellectually and each part might be well-written, but there's no emotional thread connecting the different events, nothing that makes this series of events a compelling story. There's no depth.
Of course, this is a theoretical argument; someone else might have a different opinion, and we could talk about it till the cows come home. But the proof is in the pudding, or lack of it: despite considerable time, money, and effort invested in these solutions over the last twenty years, branching storylines have not caught on. The Choose-your-own-adventure series is out of print; video games have given up their experiments with branching narrative; and nothing has come of experiments in TV and movie branching. As the interactive storytelling field now almost universally acknowledges, these many branching paths all lead to a single dead end.
Simulations, MUDs, and MMORPGs
The second approach to interactive storytelling is to create a multiplayer simulated environment, complete with places, items, and creatures, and allow people to do what they want in it. The idea is that interesting stories can arise out of the players' interactions with the world or with each other, without any intervention by a human author. This is the model for MUDs and MMORPGs such as Everquest, Dark Age of Camelot, World of Warcraft, and many, many others, as well as for more "real-life" simulations like The Sims Online. (The differences between these are not important for the purpose of this discussion, so I will lump them together and call them all "simulations").
The first thing to notice about this approach is that, unlike branching storylines, pay-for-play as well as free online multiplayer RPGs and simulations have been a huge popular and commercial success. Everquest, which started in 1999, has over 400,000 players. World of Warcraft has 1.5 million. And the players take these games seriously: the time spent by the average player is over 20 hours per week. Ebay is filled with simulation players bidding real cash for game characters or items (one player famously paid $26,500 for an island in the online world Project Entropia). Clearly, simulations have tapped into something big.
The second thing to notice is that simulations have a lot of the virtues of roleplaying games that I looked at in the last section, and have overcome some of their flaws. Of the "seven treasures," simulations clearly have at least four: the two social aspects of roleplaying (gathering of friends and teamwork), plus strategic problem-solving and roleplaying. One could argue that simulations also have creativity and spontaneous action, although for players these are limited by what the designers programmed ahead of time. As for the "five dragons," simulations have solved the limitation of location, dependency on players, and the greatest dragon of all—the limitation on the number of players. Simulations, whether based on science fiction or Greek mythology, still require significant setup time and actually longer in terms of playing time than roleplaying games, but for dedicated players this is a big step forward.
This tells me that I was on the right track: the treasures of roleplaying are attractive to a larger audience, if delivered in the right way. But I knew that simulations themselves were not what I was looking for. Why not? Because they lack the most important treasure of roleplaying: mythic storytelling.
These games might be mythic but, to paraphrase an old adage, what's interesting about simulations isn't storylike, and what's storylike isn't interesting.
What's interesting about simulations is the social interaction. The social behavior in these online RPG games is as rich and varied as in real life: players buy and sell goods, form groups, elect leaders, write laws, build homes, wage wars, and do any of a thousand other things that people do in real world communities. Real-world relationships form (or break up) based on what happens online. The social interaction is the most common reason people say they play simulations; I'm sure people could write books on why this is compelling.
In most simulations, layered on top of this community is a set of goals. Characters might be able to gain levels by killing monsters and gathering treasure, for example, or they might be given quests to explore special areas or retrieve items. These are stories of a sort, but they're not very interesting stories. They have none of the richness of relationship, dramatic themes, or emotional decision-making that makes for a great story. The players get emotionally invested in the outcome—just as a player might get emotionally invested in the outcome of any game—but there's no in-game emotional meaning, no emotional significance for the in-game character apart from that of the player.
And, to be fair, I don't think that the players are looking for that. These games are about community, and communities create their own stories: dramas about the competition for guild leadership or online romance, for example, that are significant for being personal. I think most players are happy with that. But those of us who are interested in myths—stories with epic, universal themes that have an appeal beyond the personal—will need to look elsewhere.
The AI Approach
The third approach to interactive storytelling has received the most attention and research from academia. Most of the work presented at TIDSE 2004 and other interactive storytelling conferences is based on this approach. The idea is to build software that can create a story in real-time in response to the reader's choices.
Of course, no one is trying to create a completely free-form storytelling computer. A human author would need to program the software ahead of time with a storyworld - including places, characters, dramatic themes, etc. The author might even write dialog or create art. But the author wouldn't create a complete story, just the pieces of one; the computer would figure out how to put the pieces together as the readers chose their way through the story.
When I first heard this idea, I got excited. I saw this as putting my skills as an interactive storyteller into a computer program; if this was possible, then for the first time the game experience could be shared with a mass audience. Though I'm not a computer programmer or artificial intelligence researcher, I started looking very closely at what they were producing.
And I was disappointed by what I saw. The interactive stories produced by these online RPG games are barely coherent, much less compelling. Of course, these are demonstrations of research, not actual products, so the research might be valuable even if the stories weren't. But, not being a computer science researcher, I can't really judge the quality of the algorithms, all I can judge is the results. And almost without exception, machine storytelling has produced nothing even close to an interesting interactive story.
I say "almost" without exception because there is one exception. It was presented at TIDSE, has been widely shown at other interactive storytelling and game development conferences, and is often justly cited as the best work in the field. It's a project called Facade, by Andrew Stern and Michael Mateas.
Facade is a "one-act interactive drama"—a computer program that tells a story in which you the player participate and affect the outcome. The action takes place during a get-together at the apartment of Grace and Trip, a married couple who are friends of yours. The evening begins peacefully, but things quickly turn ugly between your friends, and you're caught in the crossfire. In the end their lives and yours are changed depending on the choices that you make.
Facade incorporates a number of advancements in the art of AI interactive storytelling. The interface is a 3D graphical world: you can move, pick up objects, and gesture in real time. You converse with Grace and Trip by typing in free text, and they talk back (with prerecorded dialog), as well as show appropriate gestures and facial expressions. Most interesting from my point of view, the system is programmed to strike a balance between responding to your immediate actions and satisfying an overall story arc. This seems like a good model for interactive storytelling, and similar to what I've done as a gamemaster.
Michael Mateas showed Facade at TIDSE 2004, and he was kind enough to let me try it out afterwards. Playing Facade is definitely a compelling experience. The combination of gesture, facial expression and voice in the computer characters works well together to give them a personality; their responses to your words and actions is often surprisingly appropriate; the drama as a whole builds in a satisfying way. Facade hasn't been released yet and the version I saw was incomplete, so I can't speak to the final work, but based on what I saw I wouldn't be surprised if Facade meets its goal of being a "fully-realized, one-act interactive drama." And that would be an amazing achievement, given the state of the art.
But when I thought about using the same method to tell the kinds of stories I'd like to tell, I didn't see how it could work. In their paper on Facade, Mateas and Stern talk about the constraints that they imposed on their storyworld to make Facade possible: Facade has only two characters besides the player, the player is not the protagonist and so has only limited input into the course of the story, and the story takes place all at the same time and all in a single location. These constraints are necessary to limit the number of things that the AI has to understand, in order to make the AI possible. And in fact, even in Facade, with its tightly constrained, one-act world, the AI only understands about half of what the players say. (Though thanks to some clever tricks, it seems to understand more than it actually does.)
But what about stories with dozens of characters, and many locations? What about stories that take place over a long time with many interactions? What about stories where you play the protagonist and you can do wherever you want? Those are the stories that I want to tell. Maybe someday stories like these will be able to be told by a computer program, but given the state of the art, I don't see it happening anytime soon.
And there's something else that made me turn away from the AI approach. The goal for AI researchers is "interactive" storytelling. But what the player is interacting with is a computer program. And since the storyworld is created ahead of time, nothing can happen that hasn't already been programmed in. There's no actual interaction between the author and the audience, nothing that could create something that neither would have thought of on their own.
That's how it works for non-interactive storytelling, too, but it's different from my experience as a roleplayer. In roleplaying "interactive" means that the players were interacting with human beings: with each other and with the story author, and the story emerged out of that interaction. Stories were created that no one person had thought of.
Especially considering the great appeal of social interaction in simulations, I didn't want to give up human interactivity. I wanted to be able to tell interactive stories with—and not just to—a large audience. But how could that work? How could people interact on a mass scale and still take part in a single story?
And then I had an idea.
Discuss this article