How about a more bottom-up way of reconceiving our notions of story?

First, I'd suggest it'll be easier for the sake of discussion to not use the term "story" or "game", they have so much baggage, we all spend too much time arguing about our various definitions of them. Instead, let's simply be descriptive and concrete about what we're talking about.

Imagine an interactive scenario where the player is carrying a weapon, and there are fellow NPC comrades also carrying weapons, and we are given the goal to work together to fight an enemy squadron, in some war-like situation. Familiar design territory - it's well-understood how to make that compelling in and of itself.

Now, what if we add more verbs to the player's repertoire - for example, the ability (through some user interface, perhaps typed input, perhaps speech rec) to express a few in-domain things in natural language at any time during the game, such as concern ("this seems dangerous", or "what if we don't make it?"), encouragement ("we can do this if we stay strong"), internal fear ("I don't want to die"), hope ("I know I'm going to make it home to see my kids"), etc. Plus basic verbs such as agreement ("yeah"), disagreement, maybe, don't know, etc. Out of domain utterances ("I'd like a banana") will be essentially ignored by the NPCs; we're in a war situation, after all, it's believable to ignore unimportant stuff being said.

Think through how these new player verbs (concern, encouragement, internal fear, hope, etc.) would impact the core gameplay of fighting the enemy squadron. In many ways it would be the same as before, but now with more nuance, emotional depth, and character development. But in some ways it would be different, in which using these new verbs would affect the performance of the NPCs, and therefore become a gameplay element itself.

Is this story? Not exactly - but that's fine. It _is_ about expanding player expression, and having a deeper effect on the detailed events that occur, and the emotional tone of the experience. It'll take some real AI development to pull it off, of course. That's what we should be thinking about, not analyzing how traditional stories work and trying to match them.

Aha! You took my bait Andrew - welcome to Game/AI :)

Ok. So, now that we're talking about the *real* stuff and not the media theory, here's what interests me. How best should we model the meaning of player discourse acts for NPCs taking part in an ongoing simulation that includes general (e.g. combat AI) and local (scripted) behavior?

The obvious answer that I guess most people would suggest is "add a general model of soldier emotional state, and a general model of player discourse act identification, and then hook them in to all the other general systems." Now, I'm sure you'd agree with me that this is *not* easy to do with any depth! To really make sense in a given moment, discourse semantics must be bound to a great many contextual features. This approach implies 'marking up' vast quantities of scripted scenario in terms of a general emotional model, a nightmare for production. The icing on the cake though, is that even if you achieve all that, the discourse will be shallow and repetitive, because its forcing all emotional interaction into one general model.

But of course this is one of the things that Facade's approach attempts to solve, no? By constructing NPCs actual *behaviour code* on-the-fly from the requirements of a given beat, we don't have to give them the signals in terms of a general model, hoping that they'll come back with some appropriate response from their general behavioural competencies.

This continuation of the Oz Project's research philosophy and its opposition to parameterizing character, is what most fascinates me about Facade!

Now we're hitting on the real production problem we all face in creating deeper, more reactive behavior - how are we going to implement it effectively, and in a timely way?

It goes without saying, the inescapable fact is if developers want to achieve more depth/complexity, it's going to require *a lot* more content. This of course is a primary issue for next-gen games, whether we're talking about visuals, AI/behavior, what have you. This was the crux of GDC's most popular talk last March, Will Wright's "The Future of Content". Is procedural.

Once game developers start giving players more discourse acts (verbs), there's no way we can solely hand-author the variety of behavior we're going to need in order to handle the range and depth of contexts we'd like to support. We're going to have to move towards more generative solutions.

As you mentioned, yes, Facade builds upon the Oz Project's solutions to creating rich behavior - work within an langugage/architecture that has specific features for elegantly managing hierarchies of parallel and sequential behaviors (goals, subgoals, further subgoals), success and failure of the various subgoals, conflicts between goals, reactivity (preconditions, context conditions), etc. - originally called Hap; see Bryan Loyall's thesis. This approach gives good support for intermixing of parallel hierarchies of behaviors, a key technique for becoming more generative.

Facade adds (among other things) new features to Hap to enable more generativity, specifically, using meta-behaviors that do simple but powerful surgery in real-time on other behaviors, effectively (re-)constructing behaviors customized to the current context. We keep this simple in order to keep it sane. :-) But it's only possible because it's working on top of an already powerful language/architecture.

The updated language/architecture we built, which also includes joint synchronized behaviors between agents, is called A Behavior Language, or ABL. (Joint behaviors + meta-behaviors are the heart of how Facade's beats work.) I'm not here to make a plug for ABL, but since we're talking about the need for solutions, I'll go ahead and say that ABL will be available under a free academic-use license this fall, and is available now for commercial licensing via Procedural Arts including a library of 500 supporting behaviors, useful for narrative planning, interaction handling and discourse management, body resource management, etc. The Institute for Creative Technologies, where I'm currently doing a few months of consulting to help build prototype interactive-drama-like scenarios for training soldiers to deal with civilians and interpersonal situations in Iraq, is our first licensing customer.

The larger point is, developers need to think about building languages/architectures that do more of the work to manage the complexity of intermixing, hierarchical behaviors, as Damian put it, "ending the HFSM tyranny, and then once they have that framework, start implementing even more generative techniques on top of that...

It's a little bit difficult to have discussions about the details of how to reconceive our notions of story without specific architectures in mind to implement them. That is, it's much easier to speak in terms of specific capabilities of a language/architecture - how we would achieve desires effects within those systems...

And what's wrong with insanely-ambitious drama managers, I'd like to know?

Once upon a time, the "go anywhere, interact with anything" game was called insanely-ambitious; now it's called Grand Theft Auto. (Or The Sims, which allows a much wider variety of interactions.)

A toast to insane amibition, say I:

"But why, some say, the moon? Why choose this as our goal? And they may well ask why climb the highest mountain? Why, 35 years ago, fly the Atlantic? Why does Rice play Texas?

"We choose to go to the moon. We choose to go to the moon in this decade and do the other things, not because they are easy, but because they are hard, because that goal will serve to organize and measure the best of our energies and skills, because that challenge is one that we are willing to accept, one we are unwilling to postpone, and one which we intend to win, and the others, too. "

-- John F. Kennedy, September 12, 1962

Less than 7 years later, we did it.

> And what's wrong with insanely-ambitious
> drama managers, I'd like to know?

Only the fact that they don't work.

Sometimes insane ambitions represent actual insanity. For example, I am still waiting for the personal time travel device, aero-car-moon-lander-jet, and personal pocket fusion reactor that have been promised me.

Call me a curmudgeon, but I can't find it in my heart to blame the engineers of the world for not giving me these devices. I recognize that they would violate basic laws of physics, and I shake my fist at God and move on with my life.

> Once upon a time, the "go anywhere, interact
> with anything" game was called
> insanely-ambitious; now it's called
> Grand Theft Auto.

Despite its many virtues, Grand Theft Auto doesn't even come close to letting you "go anywhere" and "interact with anything." With all due respect, you are grossly overstating your case.

And which laws of physics do drama managers violate, pray tell?

Make no mistake: we WILL create them. The only questions are who, when, and how. But there's nothing in computing to prevent it. They're not logically impossible; they don't try to trisect the angle.

I didn't say GTA IS a go anywhere, interact with anything game; I said that's what it's called. GTA is widely characterized that way. Even if it doesn't meet your standards, it does for some people, which means it's clearly a step in the right direction.

Somewhere in between extremely pragmatic and insanely ambitious is probably where we want to be. :-) I'm sure everybody has lots of ideas for innovation - but the realities of what's sane to undertake technically, and the realities of what our employers/funders are willing greenlight, are constraints we have to live with and operate within. I'm sure we all share Ernest's underlying motivations to push forward, even if we don't all wear it on our sleeves.

This begs the question, what are realistic strategies for game developers to innovate, particularly in AI? Are the bloggers here satisfied or comfortable with the current pace of, climate for, or strategies for, AI innovation in games? More pointedly, are project-by-project incremental AI improvements a strategy that will work in the long run. The way game dev jobs are structured, can substantial progress towards, say, more fully-realized virtual characters of the nature I suggested in my first comment, be made in the long-term within the slivers of R&D time given to you during game production? I'm sure it varies from company to company, project to project, but what's your gut feeling about how it's all going to play out over time?

> And which laws of physics do drama
> managers violate, pray tell? ... They're
> not logically impossible; they don't try
> to trisect the angle.

Clearly it all depends on the drama manager in question.

I was driving home the point that some things that seem to be impossible seem that way because they really are impossible, or are so infeasible that we might as well consider them impossible. When considering some of the most ambitious possibilities for drama managers, one very frequently runs into those sorts of challenges.

Yes, many of these problems are solvable in theory, but many of them are also so far beyond our current capabilities that they might as well be impossible. Drama managers which succeed, like Facade, are ones that make good compromises with the reality of what's currently feasible.

The kind of drama manager that I know many game designers are really looking for is something that would call on so many of the capabilities of the human brain that are currently beyond our reach -- robust natural language capabilities, a massive knowledge base that represents everything a modern human actually knows, entirely procedural human animation, and human reasoning and emotion -- that for all intents and purposes, they won't be achievable until you can put the equivalent of a human brain in a box and connect it to your platform of choice. And not just any brain, but one thoroughly trained with all of the learned knowledge of a 20-to-30-year-old individual with a considerable level of theatrical talent.

> I didn't say GTA IS a go anywhere, interact
> with anything game; I said that's what
> it's called.

Sorry, but this seems like something of a reinterpretation of your original comment. Your original comment read very much like "They said it couldn't be done, and hey, look, we're already there!"

> Make no mistake: we WILL create them. The
> only questions are who, when, and how.

I look forward to that. When it happens, I'll drink a toast to John F. Kennedy from my wheelchair in my nursing home. Until then, though, I'm going to have to continue playing the curmudgeon.

------------------------------------------------

Andrew, if it's OK with you, I'm going to take your question and post it in a new thread -- partly to avoid cluttering this thread, and partly because I think it's a really important question and I'd like to call it out.

Look for the new post "Andrew Stern's question" in the main blog and post replies to Andrew's question there.

That's cool, but feel free to rename the post something more descriptive, like "Strategies for Innovation" or "Will Incremental Improvements Work?" or something like that... :-)

Done!

Actually, Andrew, I'd classify Facade as "insanely ambitious", while Fahrenheit is more along the lines of "extremely pragmatic". What do the folks here think of Fahrenheit (like Facade, a large demo was recently released), by the way?

Do you have a link for that, Jason?

It's linked here:

http://www.eurogamer.net/article.php?article_id=60240

I have played it, and it's definitely worth a little discussion. Unfortunately I don't have time to comment right now :(

I've been following Quantic Dream's Fahrenheit for some time now, it's now renamed Indigo Prophecy. Haven't downloaded the demo yet.

I'll comment a bit later this week or next too (am at Siggraph today on a panel about AI-based believable characters).

Jason Hutchens wrote:
>I'd classify Facade as "insanely ambitious"

Well, the metaphor I like to use is: when you shoot for the stars, you might hit the moon; with Facade, we think we got into orbit.

We're talking more about drama management back over at GTxA, in the comments of this post:
http://grandtextauto.gatech.edu/2005/08/09/a-few-facade-post-release-comments/