GDC: A New Vision for Interactive Stories

This is a capsule summary of Ernest Adams’ presentation from this year’s GDC, entitled ‘A New Vision for Interactive Stories’. Although Ernest operates under the banner of my company (his opening slide states ‘I’m a member of International Hobo’), I still rely upon his GDC presentations as a key means of exploring his thought process, as our regular exchanges are more focused on business issues. This year I felt he was in sterling form, and have decided to present this as my final report on this year’s GDC convention. I personally guarantee that I have misrepresented Ernest at some point in this summary, probably by interpolating and synthesising my view with his. I hope you enjoy it nonetheless!

Ernest begins by disclaiming the pomposity of his own title, noting that there is hubris in the word ‘vision’ and arrogance in the word ‘new’. He covers his modesty by stating that the content of the talk is perhaps not new in any absolute sense, but at the very least, it is new to him. 

Aristotle, Campbell & McKee

Since the conference description paints this session as an open attack on various sacred cows of narrative, Ernest is quick to explain that his actual content diverged significantly from the written description. He presents in his first slide what he calls “the Holy Troika” of narrative: Aristotle, Joseph Campbell and Robert McKee. Each is addressed in turn. 

In the case of Aristotle’s Poetics et al, Ernest observes that Aristotle's work is not necessarily applicable to games. The idea of a story having a beginning a middle and an end falls by the wayside in games, because they can have multiple endings, multiple middles and even multiple beginnings! Additionally, the three act structure works adequately for plays and films, which run for a few hours, but not necessarily for games which can weigh in at some forty hours or more.

In the case of Joseph Campbell, he is keen to note that Campbell was a folklorist. His work, while it has been adapted as a template for film writing, was never intended to be applied in this manner. Campbell never said this is how to write stories, rather he said: I have looked at mythology, and this is the common pattern. Even accepting the Hero’s Journey as a template still requires a story which is about the journey of a hero, which need not be the case.

In regard of Robert McKee, Ernest notes that he has interesting things to say, but his comments are always rooted in the assumption that what is being discussed is a screenplay, not some form of interactive content. With this in mind, its applicability to games must necessarily be limited.

He sums up this introductory segment by noting that if everyone had slavishly followed these three templates, we would not have the world’s greatest literature. Templates are useful for understanding, but they are not panaceas, and in particular, they may not be relevant in the context of interactive stories which, after all, are very different from traditional static narrative. 

Traditional Assumptions

The core experience of trying to understand interactive narrative, Ernest suggests, is: you are in a maze of twisty little passages, all different! He makes amusing reference to Dragon’s Lair, referring to it as “the decision tree of death”. [For those who have not played it, the game consists of a branching tree of actions, almost all of which result in player death. The ‘play’ of this game is in knowing which set of decisions will permit the player to survive to the conclusion]. 

He furthers this basic idea by identifying three traditional assumptions about interactive stories:

“Our goal is to create a sandbox that allows maximum freedom”: this is presented as something of a utopian fallacy, based upon the assumption that ‘some day we will be able to do anything’. Although most game designers have shared in this dream at one time or another, pragmatically such a state of affairs is far beyond our realistic or conceivable capabilities. 

“Interactive stories shouldn’t be games”: this is a difficult position to validate; after all, the scope of the term ‘game’ is in itself quite ambiguous. The feeling that an interactive story should be something apart from what we conventionally consider to be a game follows from presuppositions that need not be true.

“The player shouldn’t have to think about rules”: the idea here is that in a story space, the rules should be part of the underlying architecture, not presented to the player as part of their experience. 

In the case of all three rules, Ernest does not intend to support or deny the traditional assumptions, but rather presents them as a foundation for the rest of his discourse.

Credibility

Façade, as I’m sure most readers are aware, is a one-act interactive drama created by Procedural Arts (principally the work of Michael Mateas and Andrew Stern). Ernest notes, as an aside, that it is also (to his knowledge) the first and only Dogma 2001 game. [I presume everyone now knows this game inside out – so look elsewhere if you don’t yet know about it].

He presents – apparently tangentially – a bizarre Façade playscript in which the player has decided to play a character, Audrey, who has been shot prior to arrival at the apartment where the drama plays out. Naturally, the main characters ignore the player’s references to being shot, resulting in an amusing mismatch between the player’s text and the game’s responses. Here’s part of the transcript: 

(Audrey 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! -- (interrupted)

AUDREY
HELP ME

TRIP
Uh...

TRIP
Well come on in...

TRIP
Uh, I'll -- I'll go get Grace...

AUDREY
THERE WAS A MAN WITH A GUN

TRIP
(unintelligable arguing)

TRIP
(unintelligable arguing)

AUDREY
HELP ME I'M BLEEDING

TRIP
(unintelligable arguing)

AUDREY
TRIP

AUDREY
HELP

AUDREY
I'M GOING TO DIE

GRACE
Audrey,

GRACE
Hi! How are you? I'm so happy to see you after so long! -- (interrupted)

(and so forth. The whole thing can be found somewhere in here if you are interested).

Ernest uses this as an example of violation of credibility. Façade is unable to respond intelligently to what the player has entered, because it is being presented a situation entirely outside of the scenario it encapsulates. It posits the existence of a gunman which was invented entirely by the player, and which has no existence in the game world.

He presents an idea from Ken Perlin, which he presents as a Law (although he is quick to point out that Ken did not present the idea in this context – people seldom declare their own laws):

Ken Perlin’s Law:

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

There is, in effect, a “credibility budget” in any interactive story. (Ernest notes that Ken didn’t specify credibility as the economic context, but nonetheless, this is how he has chosen to run with it). Both the designer and the player draw upon the credibility budget. If the designer blows it, the player becomes lost. If the player blows it, they lose the designer.

For example, materialising a chicken into an interactive story space out of nothing but thin air should be an expensive operation! The designer is quite entitled to say that you cannot materialise a chicken as it will completely blow the credibility budget. 

The idea, therefore, is that a story generation system must maintain a credibility budget, and this this protects the story space from being pushed beyond its natural limitations.

Role-playing 

The term ‘role-playing’ has become practically devalued by overuse, but it still has at its heart a clearly defined context. Ernest suggests, in a stylised show of Eureka-like realisation, that Façade is a role-playing game. (But it is not a dungeon crawl). Ernest notes that in a typical cRPG you not a hero, but rather an itinerant second-hand arms dealer. Façade is not D&D but it is a role-playing game – all interactive narratives are role playing games, because there is a role you are playing.

Role-playing does not mean total freedom – it still has rules and a magic circle. When you play a game, you must accept the premise of the game. You can play a business simulation as a communist, or a wargame as a pacifist, but you will lose.

As an industry, Ernest suggests, we’ve been treating the player like the reader of a book – a tabula rasa – but in fact we are co-operating with the player to make the story. He follows this by suggesting that the lack of a requirement specification for an interactive story is a part of the problem facing the creation of interactive stories. 

We can impose laws upon the player, because the player accepts as part of the premise of play that their experience is bounded. We can impose physical laws – we may absolutely refuse to violate the physics (no materialising chickens!) We can impose social laws – inappropriate behaviour will get you locked up. We can impose dramatic laws – bad role-playing can cause the story to end. There is a balance between interactivity and narrative (a point Ernest has frequently raised in the past) which is mediated by the social contract of role-playing.

Procedural Stories 

A classic approach to interactive story is that of the branching narrative. These suffer considerably because the finer the granularity of the decisions or branching conditions, the worse the combinatorial explosion (imagine a set of ten binary choices – that’s 2^10 = 1,024 different paths!). However, this is a cost of development argument against this approach, not a philosophical problem. That said, there are other elements which are highly restrictive in this method. Time is implicit to the tree in a branching approach – events must occur in the strict order specified by the tree. Furthermore, decisions are constrained to affecting the plot, and not the characters (at least in general terms).

The opposite end of the narrative landscape are emergent narratives. To these, Ernest comments that “conventionally trained writers are not used to generating narrative events in Excel.” He contends that trying to devise the ultimate social simulator is overworking the problem – there is no need to determine the states of every character at every moment.  

Instead, we can consider a procedural approach in which situations are functions and people are parameters. As a whimsical example, the following function is presented:

function murder (victim, murderers, relatives) 

e.g. murder (King, Cladius & Gertrude, Hamlet)

By building interactive stories at a procedural level, and creating situations which are character agnostic, we have the potential to explore the problem from a wholly different angle. 

Two specific games are cited as formative steps towards this approach: King of Dragon Pass (the IGF winner from 2000), and Against the Flying Circus (by Tuonela Productions). [The core International Hobo team has also experimented with procedural narrative in a few projects which did not proceed to full development. I may try and get permission to reprint some of the concepts from these designs into this blog if there is interest].

Conclusion 

The talk comes to a close by reiterating the idea that credibility is the currency of all narrative, and that the social contract of role-playing mediates the tension between interactivity and narrative.

No other form of interactive entertainment tries to be all things to all players. Why should interactive stories have to shoulder that burden? It’s time to stop apologising for not working miracles and get on with the job of creating interactive stories.

Comments

[Patrick]

The irony is that, if Earnest had payed any attention to Chris Crawford's work, he would know what Chris has done just this. The basic unit of content in a storyworld is the reaction script, which takes a verb as its function, boolean roles as its parameters, and inclination formulas as the procedural weights which lend to different possible reactions. A good script design ensures crediblity, and the inclination formulas need to be tested and balanced to ensure this consistency. Inclination formulas are linear equations taking two variables as parameters, these variables can be picked from the reacting Actor's personality values, which have three dimensions, the Actors actual traits, the percieved traits of others and (this is where it gets complicated) the estimated perceptions others have about a third party, themselves or the estimating actor (so triangular social relations, in other words). So, thats the best I can boil it down, I think it lays things out nicely, Storytron is complex, but writing that just now actually helps me understand it more clearly. Can't wait to start building with it!

As far as characters go, Storytron could also be said to be character agnostic, characters are only defined by their p-values, which are parameters for inclination formual. However, something Crawford hasn't addressed (besides production values, unsuprisingly) is the potentials of embedded text and embedded character design to really enchance a storyworld. On this level, I'm not talking about procedural stuff, but more like what Dr. Isbister goes over in her upcoming book.

But I'm sure you can imagine why Ernest wouldn't bother looking into this thrust of development on the problem.

[Chris]

I'm pretty sure there's no bad blood between Ernest and Chris. Ernest has spoken most respectfully about Chris as one of our formost simulationists. If there is an issue there, it's not something Ernest has ever mentioned to me, and it was certainly not part of his stated motivations for this talk to comment on Chris' work one way or the other.

Your comment makes it sound as if Crawford's methods are plainly described somewhere and easy to follow. I have to say, as an observer on the outside it's strikes me as largely impenetrable right now. And it's not that I haven't spent the time to read about it.

I'm sure I speak for a lot of people when I say we're all enthusiastic to see what can be done with Storytron. But I don't think you can hold it against us that we have only a dim grasp right now of how it is going to work in practice. You, as someone planning to work with it, are in a priviledged position in terms of understanding its inner workings. I don't think you can really hold it against the rest of us that we don't have your insider perspective.

[Patrick]

I certianly don't hold it against anyone, I'm just trying to train myself to explain it as clearly as possible. But maybe its like Yoda saying "yeah, just make that spaceship rise out of the swamp, whats your problem?"

I've been puzzling over Chris' technology for about a year now, and am only now ripening my understanding enough to jump in and know what I'm doing. I really want to make it easier for people to follow the same path in a shorter time frame, so I don't mean to discuss it with an ivory tower elitism, but rather a democratized tutorial. Somebody has to, god knows Chris Crawford isn't about to do it.

I'm glad to hear Ernest has given Chris fair respect, I wish Chris wouldn't be so stand-offish, as he was at the rant. Supposedly theres a back story, but I don't want anything to do with dramas that happened when I was six.

[Michael Samyn]

Chris Crawford has written a book that very clearly explains his ideas: "Chris Crawford on Interactive Storytelling". But we'll have to wait until his Storytron software is released to really know how useful it is.
I for one have my reservations concerning the theory, but I'm eagerly looking forward to seeing the first products made with his software. If I understand it well, his company will also act as a kind of publishers for these products.

[Jack Monahan]

I enjoyed the discussion of player agency in terms of role-playing and the credibility budget. It reminds me that basically my favorite role-playing games are the Hitman series. While most cRPGs do in fact give the player the role of a second-hand arms dealer as well as kleptomaniac object hoarder, the Hitman series of games are of course very simply about being a hitman.

While that's a fairly narrow realm of occupation, and we all know it's very common territory for most games, the degree to which the games are willing to allow the player to express just what kind of hitman they want to be is what makes them interesting. For me, at least. Am I the consumate professional? The surgically precise Silent Assassin that the game rewards? A raging lunatic with a very big gun? An outcast with a fetish for costume-changing? That's all up to me.

When a game is trying desperately to cater to such a wide variety of play styles, being everything to everyone like the recently released TES: Oblivion is (which I am enjoying currently), not only is it prohibitive in terms of development cost, but it's so much harder to design around the vastly different play expectations. Being a hitman isn't something everyone wants to pretend to be in a game, but for those that do, the social contract between player and designer can work particularly elegantly... especially with such obvious and enjoyable verbs to deploy, like clothes taking and body dragging/hiding.

There are some strong indications that the next title in series will make the best of the available credibility budget by letting players play a certain way (making very sloppy hits) and responding to it in future missions (police presence heightens/hitman anonymity declines).

[Mory]

I didn't understand half of what you said, especially when you were throwing around such terms as "credibility budget". But this talk of role-playing (and the hilarious Facade script) does remind me of an idea I had a few years back.

The more control you give the player, the more rigidly you should define the character he's playing. What I mean is that before the player actually starts playing the game itself, there should be a screen briefly explaining what the character's backstory is, and what he's trying to do here.

Everyone talks about giving the player control over the story, but you never hear about giving the player responsibiliy over the story. In other words, if the player chooses to play his character incorrectly, too bad. We should allow him to completely ruin the game for himself. If he wants to see the game progress in a way that makes sense, he should play the part handed to him.

By telling the player who he is to be playing, the potential for implausibility is no longer the developer's concern. Instead, he can focus his efforts on having the possible outcomes which do makes sense improved upon. Another benefit is that it will be much easier to give the player control of various characters over the course of the game; the introduction screens would be enough to make the characters feel distinct from each other, and the player would continue to play them differently from each other.

[Chris]

Mory: Ernest spent much more time explaining his terms; I kind of just condensed his content above, so it's my failing if it didn't come across.

The idea of a credibility budget is worth explaining. Consider it this way. The player wants to connect with the game; immerse themselves, if you will. But it doesn't take much to break that immersion. For instance, if you're playing a Star Wars game and Barney the Purple Dinosaur walks in, the immersion is broken because Barney has no business being in a Star Wars game. This is equivalent to saying the credibility budget is exhausted when Barney arrives.

Conversely, a small coincidence (C3PO just happens to be sold to the son of the person who made him) uses a little bit of credibility, but not enough to exhaust the entire budget.

Some settings have bigger or smaller credibility budgets. A game set in a cartoon world of anvils and TNT has a huge credibility budget in some respects - although (for instance) a cartoon character with colonic cancer might expend the entire budget because such real world diseases have no place in a Chuck Jones-esque world. Conversely, a game set in a Jane Austen-esque period setting (I really want to make a game of Pride & Prejudice!) has a very small credibility budget. The slightest deviation from the period setting by player or designers will destroy immersion (Ernest considers immersion to be largely synonymous with suspension of disbelief).

Now when we talk about immersion, we're talking solely about the player's experience . The idea of a credibility budget is that both the designer and the player spent against a common pool of "metaphorical resource". You can imagine the credibility budget as a pile of 'credibility chips'. These get spent when the player or game designer do things that don't make sense for the narrative setting. (They are only a metaphor, though, not a tangible game mechanic).

If you dress your Star Wars character in a Philadelphia Eagles football uniform, it expends credibility beyond the ability of the game to respond intelligently to what you have done: since there is no American Football in the Star Wars universe, the game cannot be expected to recognise and respond to this outfit. It is beyond the credibility budget of the game to account for such non sequiturs. There just aren't enough chips in the pot for the game to handle this case!

Instead of a 'budget' you can think of it as just an agreement between player and designer that neither will step too far from the bounds of credibility. Small things - coincidence, for instance - can be accepted without 'collapsing credibility'. But if either the designer or the player pushes too far, then the game loses it's impact.

As a personal example, I found Kojima-san's meta-level nonsense in Metal Gear Solid rapidly used up the credibility budget (violated the potential for immersion) and after not very much time at all found myself hating it. The names in the game cost credibility to begin with - I have trouble with a terrorist in allegedly something close to the real world being called Revolver Ocelot, Psycho Mantis and Decoy Mantis. These are *silly* names. They cost credibility. There were very few 'credibility chips' left in the pot by the time I got to the point at which one defeats one's opponent by moving the controller from one port to another. This was total nonsense in narrative terms, and expended the last of the credibility budget. I completed the game only because I was playing it for research and thus obligated to do so. I failed, however, to enjoy it. This is a shame, as the game world abstraction of this game was first rate.

I prefer, as is indicated in my choice of terms above, to think of the role-playing experience as a social contract between game designer and player, or games master and player. I've written about this before, since one of my tabletop role-playing games, Contract, is built expressly around this idea. (When I talk of 'social contracts' in this post, I'm applying my terminology to Ernest's content, incidentally).

Nice to see so much commentary on this one! It's clear we're all interested in what can be done in this highly experimental field.

[Jack Monahan]

I really think Mory's got a point here. It's quite a common feature of many action games these days to have someone constantly squawking in your earpiece about your mission objectives--I found Unreal 2 to be particularly tiring in this respect--in order to elucidate your current or updated goals... but where are the descriptions of the player's intended role, his 'script' so to speak?

In a typical Unreal 2 level, wherein the stalwart space marine PC is tasked to infiltrate yet another ubiquitous space laboratory, there are a number of times the guiding voice from the mothership says "oh dear, this blast door is inconveniently locked/blocked" and advises you to find around. Nevermind the fact that this is the kind of thing that FPS players know they are expected to do from the start and at several points in the game; the point is that there's a lot of talking at the player about goals, but not about his given role.

Contrast this to the venerable game that has spawned all of these FPS games: Doom. In Doom, you could have a voice that comes on in your helmet telling you to find the blue keycard and get to the exit (as would happen in Doom 3), but the player's role was basically implicit from the start: ESCAPE. Find the exit. Fight some monsters on the way, if it suits your fancy. The player has no trouble assuming the classic one against many, lone space marine vs. demonic hordes role, because it's all pretty clear from the get-go. But while Doom is elegantly simple in that respect, there are far too many games that seem to embellish the formula without actually changing the role itself or the sense of player agency.

Which brings me back to games that very successfully hand the player a 'script', as Mory suggests they should. Hitman, as I jabbered on about in my first post, but especially the Splinter Cell series game comes to mind.
Stealth games have historically done a good job of giving the player a specific role to play, if only because stealth play was initially such a reversal of expectation, and the role of avoiding violent confrontation instead of seeking it out needed some explaining.

In its current iteration, the Splinter Cell series does a very good job of making the player feel they are super-secret, secret super-agent Sam Fisher, with some very good role-based reasons why, if they want to stay "in character", they shouldn't go killing everyone in sight, and indeed such an action might end the mission prematurely. Though stealth games are clearly not everyone's cup of tea, what seem like play restrictions actually heighten the immersion and are found to be highly enjoyable by those that play stealth games.

[Chris]

There's much to be said for being handed a role in this way, but it's not appropriate for every audience.

Over the years, "role-playing" as a term has diverged to take upon two seperate meanings.

In the first case, the Western meaning of "self-expression" creating a character (as an alter ego) and playing for escapism and wish fulfilment. Rarely very sophisticated, nonetheless, the escapism of this flavour of role-playing has a very wide appeal in the West.

In Japan, however, and among a cohesive but minority audience in the West, there is also the meaning of "role-playing" as "playing in character". The great example of this in some respects is Shenmue. The player is not given the freedom of expression expected in a Western cRPG (for instance), but instead is expected to play the role of Ryu provided to them.

In Japan, "playing in character" is more important than "self-expression" - a cultural distinction, no doubt. In the West, "self-expression" is (by a factor of about 3:1) more important than "playing in character".

Of course, there is an obvious caveat to this: if the player doesn't want to "play in character", they can always ignore their character description. This suggests at the very least the manual for a game should include a one-page "character guide" for players who would benefit from this guidance.