A New Vision for Interactive Stories

Ernest W. Adams

2006 Game Developers' Conference

This is an approximate transcript of the text of my lecture at the Game Developers' Conference on March 24, 2006. I present it in this form because the nature of the material does not lend itself to the traditional paper format. Also, because the lecture is informal and to some extent ad-libbed, this is not a verbatim document.

Part I: The New Vision


Hello and welcome. This is "A New Vision for Interactive Stories," and I'm Ernest Adams.

I need to begin with a disclaimer. I realize that there is hubris in introducing anything as a vision, and that there is a risk in introducing anything as "new." I haven't read every work on interactive narrative that has ever been written, so many of these ideas may have been heard before in other places. Rather, what's new about what I'm going to say is that it's new to me. In the course of the last year I have changed my thinking about some of these ideas rather sharply.

I'm going to start by introducing a few old friends:





Joseph Campbell

Robert McKee

Aristotle, Joseph Campbell, and Robert McKee—the unholy troika of storytelling authority. There has been a tendency in recent years for people in the game business who are interested in interactive storytelling to rush out to gurus about storytelling and to adopt their methods rather slavishly, and I don't think that's a good idea. I don't think that Aristotle necessarily works for what we're doing. His notion that every story has a beginning, middle, and end? No. We might have multiple endings, which is not something that Aristotle talks about. We definitely have multiple middles. The player can save and reload in the middle of a story and create a new middle, if he wants to. We might even have multiple beginnings, if the game is randomized every time you play it.

The kind of structure that Aristotle talks about is not necessarily appropriate for what we're doing, if we're talking about genuinely interactive narratives in which the player's actions change what's going on in the plot. The three-act structure that Aristotle was talking about—setup, confrontation, resolution—was designed for plays. He was talking about drama on the stage, and it works for movies too, because movies are about the same length as plays. But it has nothing to do with an entertainment form that can last 40 hours, like a big video game. Nor does it have anything to do with an entertainment form that can last indefinitely, like a soap opera.

Now there's a great deal of interest in Joseph Campbell, and I do think that the Hero's Journey is useful for the kinds of games that have a hero, a linear story, and a journey. But Campbell is descriptive and not prescriptive. He was a folklorist, not a creative writing teacher. Campbell never said the Hero's Journey was the right way to create stories. All he said was, this is the way a lot of stories about heroes get written. But a lot of stories are not about heroes. "The Tale of Ichabod Crane" is not about a hero. It certainly isn't a Hero's Journey.

As for Robert McKee, he has a lot of interesting and useful things to say about creating emotion, but I'm not terribly impressed with his stuff about structure, because again, he's not talking about interactivity. He's talking about screenplays. Robert McKee assumes that his audience is writing for the movies. I've never seen him claim that he knows what he's doing with interactive entertainment.

In McKee's workshop, he discusses Casablanca extensively, and he explains why it is such an excellent story—which is quite true. And eleven years ago I gave a lecture at GDC in which I also discussed Casablanca as a positive example of storytelling. But in that lecture, I said that its incredible strength as a story, its tightly-knit fabric, is what makes it unsuited to interactive entertainment. It ill tolerates any fiddling.

I think if everyone slavishly followed the templates that have been devised around these three men and their philosophies, then much of the world's great literature would simply never have been created. Not all of the great stories of the world follow these templates, and I don't feel as if we are under any obligation to do the same. Unless you are intentionally writing a linear story that fits into the format that they are talking about, set them aside.

Three Traditional Assumptions; or, How We Got Into This Mess

I've been thinking about this stuff for a long time, and sometimes it just seems as if my head is full of molasses. There are all sorts of complicated interlocking issues; here are all sorts of models from which to borrow. There are books, plays, and movies in the non-interactive media; there is the huge plethora of existing games, from Dragon's Lair's game tree of death to the near-total freedom of The Sims. Both are computer games and both are interactive narratives according to some definitions. Then there's all the material that has been written about games, and the colossal quantity of material that has been written about writing itself. How am I supposed to get any synthesis out of all this stuff? A great deal of it is contradictory. It's just really hard to think through, and it has been painful and awkward to make any sense of.

In the course of that time I have made a number of assumptions. I'm going to talk about these assumptions and how they have influenced my thinking. But they aren't just mine; I think these are assumptions that a lot of my colleagues share.

"Our goal is to create a sandbox that allows maximum freedom to the player."

The earliest computerized interactive stories were text adventures. They used typed input commands. They told the player that he was in such-and-such a place, and presented him with an input prompt. But they didn't list the commands that were available. The text adventures pretended that the player could do anything. Of course the player realized, five minutes in, that that was false; he couldn't really do anything, because the machine didn't understand very much. But the immediate reaction of anybody who played the original Adventure was, "OK, well, you should be able to do anything," and, for those of us with an optimistic attitude, "Someday, we will be able to do anything." I adopted this assumption without really thinking about it, and think a lot of people did.

"Interactive Stories Shouldn't be Games"

Our second assumption was that we should abandon the "game" concept in the context of interactive storytelling. And in fact, four years ago, I gave a lecture here called "Why We Shouldn't Make Games." I said that we shouldn't make games for a couple of reasons. One had to do with cultural credibility. The term "game" implies light, fun, meaningless, and temporary entertainment. If something doesn't matter, it's "only a game", not an art form. Stories are more important than games. Games are made by nobodies like us, while stories are made by de Maupassant, Chekhov, Virginia Woolf, and J.D. Salinger. So to broaden our medium and gain cultural credibility, I made this assumption that whatever interactive stories will be, they won't be games.

There was also a reason to do with mechanics: The whole thing about winning and losing, the struggle for achievement, all seems wrong in the context of a story. We also made a false analogy between gameplay tension and dramatic tension, assuming that they were the same kind of thing when they're not. I addressed that last year, so I won't go into it again this year.

So that was another basic assumption: Interactive stories should not be games, or they won't be games once we get them figured out.

"The Player Shouldn't Have to Think About Any Rules"

The third assumption arises from something that we have taught the players over the years. In an ordinary board game you have to obey rules, and to obey them you have to know what they are. This is a conscious process. There is a list of permitted and prohibited actions, and you are aware of the rules at all times. But video games hide the rules. This is great, because it contributes enormously to player immersion. The game knows the rules, so you don't have to. The permitted actions are implemented by the user interface, and the prohibited actions are simply not available. So we have trained our players to believe that if a thing is possible, it must be permissible. If they're not supposed to do a thing, it shouldn't be available.

So I want you to keep these assumptions in the back of your mind, because I'll be coming back to them later.

Façade and Its Critics

Now I want to talk about Façade. Façade came out earlier this year. Very briefly, it's an interactive drama in which you play the role of an old friend of a couple, Trip and Grace, whose marriage is on the verge of a breakdown. You've gone over to their place for drinks, not having seen them in a long time. You can walk around their apartment, touch Trip and Grace using the mouse, and above all, talk to them by typing in text. They understand you – most of the time – and they talk back, using spoken English.


Trip and Grace of Façade.

It was made by Michael Mateas and Andrew Stern, using their own money, and it's available free at www.interactivestory.net. (Incidentally, for those of you who remember what this means, Façade is the first, and so far the only, Dogma 2001 game.)

I think Façade is one of the most important computer games of all time, and the most important game of the last ten years. It is an interactive drama about a specific situation, and the story is a combination of emergent and embedded material. Parts of it are pre-scripted, because Trip and Grace's responses are recorded audio. There is no stated goal, but there are several possible outcomes depending on how you behave. If you come in and you're very aggressive or obnoxious, they'll just kick you out. But you can also cause them to reconcile or to break up.

As you play Façade, occasionally it produces absurd results. Façade allows you to keep a transcript of everything that is said, and here is one of the transcripts that I found on the Web. [The transcript was generated by Duncan Fyfe.] The player's name is Audrey:

    (Audrey [the player's character] knocks on the front door.)

    (Trip opens the front door.)

    TRIP: Audrey!!

    AUDREY: Trip, I've been shot!

    TRIP: Hi, it's so great to see you!

    AUDREY: Help me!

    TRIP: Uh, well, come on in… Uh, I'll… I'll go get Grace.

    AUDREY: There was a man with a gun.

    TRIP and GRACE: (Unintelligible arguing)

    AUDREY: HELP! I'm going to die.

    GRACE: Hi! How are you? I'm so happy to see you after so long!

    AUDREY: For God's sake, I'm bleeding…      (etc.)

Obviously, Trip and Grace don't understand what the player is talking about. People make fun of Façade because it doesn't react correctly to every possible input, and my first reaction on reading this was the same: the AI is just not up to the job. But after a while I began to realize that there's actually something more interesting going on here. I'll come back to this too. I realize I've pushed a couple of things onto the stack already, but I promise I'll wrap it all together by the end.

Ken Perlin's Law

Last November, I went to the Virtual Storytelling Conference in Strasbourg, and one of the keynote speakers was a guy named Ken Perlin. If you don't know who Ken Perlin is, he's a professor at the Media Research Lab at New York University. He's one of these guys like Chris Hecker, who's both terrifyingly intelligent and seemingly blessed with boundless energy. Chris Hecker is sort of what you get if you cross Albert Einstein with Tigger, and so is Ken Perlin.

Ken was talking about some of the various things that he's done, and almost as a throwaway remark, he said something that really brought me up short. Now, Ken is too modest to have called it a law and named it after himself, but I think it deserves it, so I'm going to do it for him. Ken Perlin's Law is this:

The cost of an event in an interactive story must be directly proportional to its improbability.

And at first I thought, what's he talking about? Adventure games don't have an internal economy, they don't keep track of costs. And in role-playing games, improbable events are just good or bad die rolls, there's no cost element. And what unit would this so-called cost be measured in?

But the more I thought about it, the more I began to realize how much sense it makes. The unit of cost of improbable events is their credibility. In fact every story, interactive or non-interactive, book, movie, television, or computer game, has a credibility budget. Ken did not say that the unit is credibility; he wasn't specific about the cost. That's my addition, so if you think it's completely bogus, blame me, not him.

A story can only tolerate a limited amount of improbability before the credibility budget is exhausted, and the story is ruined. In the case of non-interactive narrative, the author controls and spends the credibility budget, and when the author blows it, she ruins her story. In the case of interactive stories, however, the designer and the player both spend on the credibility budget. If the designer blows it, then he's lost the player. But if the player blows it, then he's lost the designer. He's done something so improbable that the designer didn't budget for it.

The example Ken used was, suppose you're playing along in an interactive story that's set in the modern day, without any magic or anything, and the player decides that he wants to materialize a chicken out of thin air. Ken said, this should be a very, very expensive operation. In my terminology, it blows the credibility budget. And the designer is entitled to decide that you simply can't materialize chickens in his world, because the credibility budget doesn't stretch that far.

In papers on interactive narrative it's very common to see grand statements of the form "the designer and the player collaborate to create the storylike experience" without any explanation of what the hell that really means or how it's supposed to take place. I don't  know what the hell it really means either, but I think this business of both the designer and the player making withdrawals from the same credibility budget when they do something improbable, is central to this idea of collaboration between player and designer. It's where the rubber meets the road on the Problem of Internal Consistency, which is one of the three problems for interactive storytellers that I introduced at this conference 11 years ago.

I'm not just talking about this stuff in an abstract, theoretical sense. I'm talking about design and coding. I think it's quite possible to build a quantity, a resource, into a game that is an amount of credibility, and to track it. In fact, I think a story-generation system, if we ever create such a thing, must keep a credibility budget. If it doesn't, it's going to generate nonsense.

The New Vision: Storytelling Games as Role-Playing Games

So what does it mean if a storytelling game has some kind of an internal economy?

In his book [Chris Crawford on Interactive Storytelling], Chris Crawford makes a brief reference to scoring systems for drama. He uses it like a classic game score, to reward the player for doing that the author wants him to do. His example is giving somebody who is playing Juliet a higher score for committing suicide at the end, because that's more dramatic, than just for walking away from Romeo's body. If the player, as Juliet, says, "Eh – teenage guys are a dime a dozen," she doesn't get the bonus points. But if she says, "Oh, happy dagger, this is thy sheath; there rust and let me die" – extra points for killing yourself.

It's an interesting thought, but that's as far as Crawford takes it. It's like a Wikipedia stub in his book, it needs further elaboration before we can judge it. But thinking about what Ken had said—the idea that interactive stories can and perhaps should have an internal economy—it struck me like a flash of lightning, what the implication of this really is.

You have to understand—I've tended to think of interactive stories in terms of adventure games (which lack an internal economy), because they're the ones with the deepest characters and the richest plots. They're the ones that seem the most story-like and the least game-like. They don't have a lot of numbers. They don't give the player an artificial goal to shoot for, and they're not about winning and losing, and so on. And that's kind of where my head has been at.

This realization was: Façade is a role-playing game. It's not a dungeon crawl, but it is a role-playing game.

Now I've been running away from role-playing games even faster than I've been running away from games in general. Computer RPGs are the ideal example of what a story-like experience is not. They have all these numbers. They're full of repetitive combat, and buying and selling weapons. As I've said before, in a computer RPG, you're not a hero, you're an itinerant second-hand arms dealer. They have this whole leveling-up mechanism that occupies most of the player's attention, rather than the plot, which is usually fairly thin.

But that's not role-playing. That's Dungeons & Dragons.

D&D is only a subset of role-playing. And it's very well-suited to computers precisely because it has all these numbers. But as we all know, D&D as played on the computer, doesn't actually have a lot of role-playing in it. Façade isn't D&D, but it's still a role-playing game because it's about playing a role in a specific situation.

So, does the fact that Façade is unable to respond to certain inputs make it a failure? No. When people make fun of it, they're assuming Façade should try to be universal. But Mateas and Stern never claimed it would be universal. Now I'm not here to defend Façade; Mateas and Stern are big boys, and they're more than capable of doing that themselves. What I am saying, however, is that those criticisms are off the mark.

Role-playing does not mean total freedom. Role-playing games still have rules and a magic circle. Going into Façade and saying, "I've been shot!" is just bad role-playing. It's like going into D&D and saying, "Hey, did you catch the space shuttle launch on TV last night?" The world of Façade has no guns in it, just as the world of D&D has no space shuttle in it. So it's no wonder that Trip and Grace don't understand when you say "I've been shot." It's not that it's bad AI. It's that guns are outside of the game world.

The Traditional Assumptions Violated

This realization of mine—that storytelling games are role-playing games—violates the traditional assumptions I described earlier:

    "The goal is ultimate player freedom." Maybe that's not tenable in role-playing.

    "Interactive stories shouldn't be games," or that when we get them right, they won't be games. Maybe games bring a beneficial structure. Maybe they require the player  to behave in ways that are consistent with the storyworld. What if I play a war game as a pacifist, or a business game as a communist? I will lose. When you play a game you must accept the premise of the game, and there is no reason why an interactive story has to be workable for a player who refuses to accept its premise.

    "If you can do it, it must be allowed." That's not tenable in social contexts. It's OK for actions involving physical activity, and we can place limits on the user interface to restrict player actions in a physical context. It's problematic when the action is speech, because we can't impose limits on what players can say. We've become very well aware of this in MMOGs, because a lot of players come into MMOGs bringing with them the same kind of expectations that they have about single-player games, namely, "If I can do it, I'm allowed to do it." And in fact, MMOGs have had to impose explicit rules that players obey voluntarily, restricting their speech. MMOGs violate the "If you can do it, it must be allowed" assumption.

Interactive narratives are role-playing games simply because they are about playing a role. That is the new vision.

Some of you may be saying, "Big deal, I got there years ago" or "Role-playing? What the hell kind of a breakthrough is that?" But to me, it finally enabled me to get through the molasses in my head, to put all this stuff into a conceptual framework that I was able to work with. And along the way it forced me to abandon these cherished notions, these wonderful dreams I had about interactive storytelling, that had actually been holding me back. I had to let go of the utopian universal sandbox. The notion that an interactive story should be free from any internal economy. Ken Perlin helped me to see the value of including one. The idea that the player had no obligations to the story. We've been treating the player like a reader of a book. Like a person we know nothing about, who doesn't owe us anything. He should be able to do what he likes. And that's wrong. Because he's collaborating with us to create the interactive experience. And that means he's spending our credibility budget.

Once I accepted that, then a lot of other stuff dropped into place. For one thing, it provides a solution to what I call "the screwing-around problem." Screwing around is a style of play. It's free-form, chaotic, and largely unbounded by rules. It's an outgrowth of the ultimate freedom assumption, and the "if you can do it, it must be allowed" assumption. It is in fact a classic masculine style of play, that has driven every little girl, who ever had a brother whom she was trying to include in her story-like game, mad with frustration. Because she wants to create a coherent experience with characters playing roles, and her little brother wants to screw around.

It's no surprise that Grand Theft Auto is lauded to the heavens by the largely male group that play it: is the ultimate enabler of masculine screwing around. Driving like a maniac, performing random acts of violence, and having meaningless mechanical sex. I can't think of a clearer example of screwing around than that. But while you're screwing around in GTA, the story is stopped! It's compartmented to prevent the player from damaging it. While you're screwing around in GTA, you cannot affect the plot of the story. They keep it separate, to let you screw around in one place and have the plot someplace else.

So showing up at Trip and Grace's door in Façade and saying you've been shot, when the game has no internal conception of being shot, is also screwing around. Role-playing places limits on screwing around.

The Laws We May Impose

One of the problems with interactive storytelling is that it lacks a requirements specification, and I think a lot of the confusion and disagreement arises from that. What are we actually obliged to provide to the player? Must an interactive story enable the player to do anything whatsoever, including screwing around? I would say no. Apart from the problem of having the resources to present "anything" (text adventures can, but not anything else), a story engine cannot handle the implications of absolutely any event. And the interesting thing is, a human storyteller can't either. If any of you have been Dungeon Masters, I'm sure that you, too, have been driven mad by your party screwing around.

What limitations may we place, then? Well, I think there are three: The physical, the social, and the dramatic laws of our storyworld.

The physical laws: The player must act in conformance to the laws of physics of his world. We may absolutely prohibit (or rather, decline to implement) actions that violate them.

The social laws: The player must act in conformance to the social laws of his world. If she violates those norms, the game is entitled to misunderstand her, to ostracise her, to lock her up as mentally ill, or to execute her – just as the real social world does.

The dramatic laws: The player must act in conformance with the role that he as agreed to play. He must accept the premise of the game, or our obligation to provide him with a coherent story is at an end. If the player screws around, all bets are off, and it's not our fault. Requiring that the player actually play a role within the context of the story enables us to place expectations upon his performance.

In other words, we can mediate the eternal tension between interactivity and narrative, between the designer's desire and obligation to construct a coherent story and the player's desire for freedom, through their common agreement that the player will be playing a role. If we try to create interactive stories with the assumption that every interactive story must be the ultimate sandbox that can handle any possible thing the player wants to do, we are setting ourselves up to fail!

We are allowed to say, "No materializing chickens!" And we're also allowed to say, "No pretending to have been shot, when there are no guns in the game world." That is the understanding that dawned upon me this year. We have obligations to the player, but the player has obligations to us, through his participation as a role-player.


I have been using this image for a long time to illustrate the tension between the player's desire for interactive freedom and the designer's desire for narrative coherency. But only recently did I realize that role-playing is the fulcrum of the balance, and add that text to the picture.

Part II: Practical Approaches

Now, how do we actually build it? The second half of this lecture will be about pragmatic issues.

Branching Trees

branchingWe're all familiar with branching narratives—the whole issue of branching tree structures and the combinatorial explosion. And the smaller the granularity of decision-making, that is, the more frequently you make decisions in the game, the faster the tree explodes. And the larger the number of options at any given decision point, the more branches there are available, which makes it explode also.

In game design, we ordinarily consider that both of these are a good thing: small granularity, frequent decision-making, is good; and offering the player many options is also good.

But I don't think the cost implications of the combinatorial explosion is the real weakness of the structure. That's a financial problem, but not a philosophical problem. The designers of text adventures don't have to worry about it as much as designers of graphical adventures, because they don't have to create so much content.

But there are some philosophical weaknesses to the branching tree structure as well:

Philosophical Weakness 1: One is that time is hard-coded into the tree. Situations must occur in a particular sequence that's built in. All cause-and-effect relationships that can occur in the course of the game's story are fixed in the tree's structure. That's OK because it guarantees that related events will occur in their proper sequence. You won't get absurd results like effects happening before their cause. But unrelated events are also hard-coded into the tree. So they have to occur in the sequence the tree dictates, even though there's no reason that they couldn't occur in a different sequence. If you want to allow for events to occur in a different order, you have to have more branches in the tree.

branching2Philosophical Weakness 2: The second weakness is that the consequences of all actions by any character (player or NPC) are hard-coded into the tree as well. The game treats decisions as things that move the plot along a tree, not as things that affect characters. And indeed, the early games were extremely plot-oriented, and not very character-oriented. So if someone lives, you follow one huge branch of the tree, and if someone dies, a different huge branch, one that contains no further reference to him. You have to have whole separate subtrees, one with and one without anyone who can die in the course of the game. Every critical decision divides the tree into these huge branches, because otherwise you would get absurd results. A person who has died must never reappear in the story, and that requires a separate storyline. So what the industry has done is to abandon branching trees and say, we'll just make rail-shooters and linear narratives instead.

Social Simulators

But in the research community, what some people have done is to go to the opposite extreme and say, OK, no more hard-coded plots, no more branching trees. All interactive storytelling must be emergent. Plots must arise by themselves as a property of the algorithms that define the relationships among people. So what we have to do is create the perfect social simulation engine, that can enable us to define characters with any degree of complexity, and that will take the relationships among these characters and generate a story from them.

OK, right. Perfect social simulation engine that automatically generates credible, coherent, and interesting plots. No problem there, then.

I think this is simultaneously barking up the wrong tree while throwing the baby out with the bathwater, if you'll excuse the expression. As I mentioned last year, in his GDC lecture in 2000, Marc Leblanc pointed out a number of problems with purely emergent narratives, and I won't go over them again. To them, I would add that conventionally trained writers are not used to doing their work in Microsoft Excel. They're even less used to doing it in code.

A pure social simulator treats life as a bunch of characters just bouncing off each other in a sort of Brownian motion. But that's not all there is to life. Life is also full of external events that intrude and place pressure on the people. Dramatic events. Plot-like events. Story-like events.

So, there's another question for our requirements spec. Must we simulate the personality of every character in detail at every moment? Well, I don't think so. First, not every character; second, not at every moment.

Where did we get this notion that every character in an interactive story must be a fully-realized human being at all times? Was it from watching Captain Picard on the holodeck? Do you think he really expected to be able to have a meaningful relationship with every single character in one of his Dixon Hill holonovels? I don't think he did.

In other media, there are loads of minor characters who don't have any emotional depth at all. Doormen, hot dog vendors, taxi drivers, receptionists… books and movies are full of people with whom the hero has one or two little interactions and then the character goes away and isn't seen again. Authors don't crank up the full power of their character-creation skills to include these people, and we don't need to crank up the full power of a character personality and social behavior simulator just to take a delivery from the Fedex guy.

And if you insist on falling in love with the Fedex guy, and the story is not about your relationship with Fedex guy, then your love is going to be disappointingly unrequited, and that's not the game's fault. I know that in the real world, bartenders are real human beings and they ache and yearn and shop just like the rest of us, but in the context of fiction, a bartender is just a bartender.

Nor do we need to implement the full power of social simulation at every moment. In other media, when characters other than the protagonist are off-stage, they aren't doing anything. And the author is not trying to keep track of everything that everyone is doing at every moment. It really is a stage. The reader's attention is a stage, and the characters who are not on the stage are just sitting around in the wings reading the newspaper.

In fact, I believe the notion of the ultimate social simulator that can handle any interaction between people is every bit as much of a pipe dream as the ultimate player freedom sandbox that can incorporate any player action into an interactive story. I think trying to devise the ultimate social simulator is again setting ourselves up to fail. It's asking us to do vastly more than the greatest storyteller of all time—Tolstoy, Homer, choose your favorite—ever had to do.

A Hybrid Solution

I think there's a hybrid solution that doesn't lock conventionally trained writers out of the process.

situations_and_peopleThe early games, with branching tree structures, subordinated all character to the plot. The result was predictably shallow characters. This, in effect, produced the linear or near-linear adventure games we know today. We have long assumed that player decisions should drive the plot and that the other characters in the game world are rather static. This is how all the classic adventure games worked.

Social simulation engines, by contrast, subordinate all plot to the characters. Everything becomes about character interaction, whether or not anything interesting is happening. This in effect produces a kind of dramatic sandbox, a simulation of personality interactions, without any forward movement.

I believe they must be balanced so that each influences the other. The approach I prefer assumes that situations drive character transformation, and this produces a loop: situations stress people, people act to change the situation.

Somebody is already putting it to work.

King of Dragon Pass

King of Dragon Pass is a management and strategy game with a strong role-playing element. It was one of the winners at the Independent Games Festival in the year 2000. In the game, you're looking after a clan of people that is governed by a council of elders called the Ring. The kind of advice you get from the Ring depends on who's on it. From time to time you can send members of the Ring as emissaries off to conduct negotiations with other clans. The outcome of those negotiations depends in part on who you send.

KoDP 2, large

A diplomatic event in King of Dragon Pass, with the current members of the Ring shown at the bottom.

In other words—and I've talked to David Dunham, one of the authors, and he confirms this—situations are functions , and people are the parameters to the functions. You put different people in to a situation, you get out different results. And these can chain onwards to produce new situations. For you programmers, the characters are pass-by-reference parameters, not pass-by-value, so the functions, that is the situations, not only change the state of the plot, they change the people as well.

This helps the combinatorial explosion problem because the characters are not hard-coded into the plot. If a character dies, you don't have to have a huge branch of the plot to deal with it. That character is simply not around to serve as a parameter to any future situations.

function murder (victim, killers, relatives);

(King, Claudius & Gertrude, Hamlet);

The situations are character-agnostic. So instead of "What should Hamlet do about Claudius and Gertrude having conspired to murder his father?" it becomes: "What should a young man do when his father's ghost tells him that his mother and uncle have conspired to murder his father?" The outcome depends on the characters involved! How he feels about his mother, how he feels about his uncle, and how he feels about ghosts. If the father was a tyrant, maybe it's all for the best. If the young man is completely terrified of the mother and the uncle, maybe he'll run away. And if the young man is really conflicted, you get three hours of waffling followed by a bloodbath.

Again, I'm not just speaking theoretically. King of Dragon Pass contains a proprietary interpreted language designed specifically for this purpose. It literally codes up situations in which characters may be placed, and the game contains hundreds of them.

Now, other RPGs also include character-agnostic situations, because they don't know in advance who's in the player's party. But the situations are almost always about clobbering something, so it doesn't really matter who's in the party. They aren't social situations, they're clobbering situations. So regardless of who you take into the party, the bad guy ends up dead. In other words, conventional RPGs could do this, but they don't live up to their potential for it. That's not the way most of them are designed.

This mechanism avoids a number of the weaknesses of the branching tree. The combinatorial explosion can be reduced because not all the actions influence every other action. Some situations arise in consequence of others (the main story arc), but others arise independently. Decisions and events can affect the future not by hard-coded chains of cause and effect, but by affecting the qualities of the characters involved—that is, the attributes that describe them.

So some future possibilities get closed off not because the player has gone down a particular branch of the tree, but because the character of the individuals involved doesn't permit it. So suppose you have a set of situations you have created that can occur in the context of a marriage. Those situations simply will not happen to characters who are unmarried. If you put an unmarried person in as a parameter to one of those situations, the function will return a null result. But you don't have to have one branch of a plot tree for married people and a different branch for unmarried ones.

Against the Flying Circus

There's another game using this mechanism as well. I'm an consultant to a business incubator named the Environment for Lucrative Virtual Interaction in Oulu, Finland, and there's a startup company there that I'm advising called Tuonela Productions. They are working on a game called Against the Flying Circus, which borrows directly from King of Dragon Pass. King of Dragon pass was mostly a management simulation about looking after your clan, and there are a lot of numbers in it that the player can see, about how many cows and sheep you have, and so on.

Against the Flying Circus is about commanding a squadron of Allied pilots during World War I, going up against Baron von Richtofen and his squadron, the Flying Circus. But it's not about flying the planes. It's about being the commander, and it concentrates on the human face of the war. Each of your pilots is an individual, and as the commander, you have to decide who to send out on which missions, and to balance the needs of the war with the current state of your squadron: which pilots are best, which ones are tired or injured, which ones need more practice, and so on.

But in addition to their missions and their state of physical health, each of these pilots has a personal life. So they can go into town, get drunk, land in jail, and not be available to fly later missions. They can fall in love, they can get bad news from home, they can fall out with another pilot over a gambling debt, and so on. All these things affect their concentration and their ability to fly, and of course, sometimes they don't come back from a mission. Then you get rookie replacements, and the story carries on without them.


And these really are the challenges that a squadron commander would face. You have to decide how to manage these guys, how to maintain good relations with the townspeople, how to make sure that your men are in good condition and mentally sharp, and of course try to achieve your missions for the war, all at the same time.

In this structure, unrelated events—Bob gets a promotion while Jeremy gets a Dear John letter from his girlfriend—can occur in any order, without having to be hard-coded into a tree-structured plot, which makes the game more replayable. But it's still going to feel very story-like, because the situations that arise are meaningful and dramatic. I think this game has considerable potential to raise the bar on the emotional depth of video games, because you develop a personal relationship with every one of these guys, and any one of them can die on any mission. Now, that potential is already in existing computer role-playing games, but most of them are so tied up with combat stats that they don't take advantage of it. The human element is left behind in most commercial role-playing games.

What I like about Against the Flying Circus is that, unlike a pure social simulator, it still gives conventionally-trained writers an important role. They can think up the situations that might arise, and they can think about how different kinds of characters would react to those situations, and in addition, how the situations would change those people. This is what conventionally-trained writers are good at, and that's what we need more of in this industry. And then they can work with somebody to turn the situations into functions and the characters into data structures.

I believe this hybrid of human-created, embedded, but character-agnostic, situations and emergent individual characters who act within those situations as their own personalities determine, is an exciting new avenue to explore. I think the hybrid is better than either of the approaches at the opposite ends of the scale.


I believe that credibility is the currency of all narrative, interactive or otherwise. I think that role-playing mediates the tension between interactivity and narrative. And I think that treating interactive narratives as role-playing creates a contract between the designer and the player, such that:

    The designer promises to provide a credible, coherent story if and only if…

    The player promises to behave in credible, coherent ways.

...and if they don't, all bets are off.

Furthermore, I think that the combination of human-designed, embedded, but character-agnostic situations, plus a character simulator, offers:

    More flexibility than hard-coded narratives and characters.

    More interesting dramatic possibilities than a pure social simulator.

… and it merits further study.

No other form of interactive entertainment tries to be all things to all players. Why should interactive stories take on that burden? I think, in fact, that we—or at least I—have been staggering under the burden of these assumptions of this colossal thing that we are expecting to try to deliver:  this ultimate sandbox, this experience whose premise the player doesn't have to accept. It's time to stop apologizing for not working miracles, and get on with the job.