Sign up for news

At AIIDE a few weeks ago, I was impressed with the number of interactive storytelling presentations. Granted some still occasionally fell into the vein of “and here the AI will calculate how to tell an amazing story”, which seems like an at least partially flawed approach.

There’s been a number of games in past year or so that have dipped their toes in the simplest water of this sort, giving us simple good and evil choices, letting us decide which end of the spectrum we want to progress towards. There’s one word that comes up pretty frequently in conversation with story-minded folks to describe their dramatic pacing – muddy.

If you can go immediately from good act to evil act, and back again, in terms of the overall narrative, there’s no real meaning or arc to the switch. It takes tremendous, incredibly rare, crafting of narrative options that the player can pursue to make sure such switches carry any dramatic weight. Unfortunately if you’re trying to give the player that spectrum of choice, you have to include those options at as many junctures as possible, regardless.

The most apparent problem in creating these narratives is that the player has to have some storytelling ability to create an arc that’s satisfying to them. People often think this is the hurdle, and this is the problem drama management AI is meant to solve. I think this is a latent ability everyone has, if properly poked and prodded. The human being is a storytelling creature.

The real problem is the player simply does not have the information available to decide what would be the most interesting arc. Obviously you don’t want to spoil the entirety of the story by laying out to the player what all their future options will be – but without that foreknowledge, the player can’t pick the most dramatic moment to make certain decisions. If they want to play the good character that tragically falls, they may betray their friend (or what not) too early, such that they miss an opportunity for a much more dramatic backstabbing. I think that’s more a failure in the interactive storytelling, and less a failure in the player’s ability.

So the purpose of “AI” here is not to pick which future options the computer thinks are the most dramatically satisfying. It’s an impossible task even if we asked the player what type of dramatic arc they wanted – they might not consciously know, and it could turn out they’d be more satisfied by something they didn’t realize. Instead, there’s a need to procedurally encourage and discourage immediate narrative options in order to hint to the player that a better payoff will come later if they delay that key decision (or make it now, as the case may be).

But that’s a topic for another post. The reality of the situation is we don’t have a lot of good techniques or technology to do that – however there’s still plenty of stories to be told via this simple kind of meandering through positive/negative space. Stories where the meaning is derived not from the dramatic changes of which side you have picked, but the indecision and change the character is meant to experience. Stories that, by their very nature, have muddy aspects to them.

Kurt Vonnegut made a wonderful analysis of Hamlet in this vein in Man Without A Country. He points out the standard mapping of certain stories as they change from the characters’ good fortune to misfortune. What some would argue makes for good story structure, the classic up and down arcs, doesn’t fit for a story like Hamlet. Vonnegut notes that each of the major events in the story could be interpreted as good or bad from different perspectives, but you can’t really say they are one or the other definitively. Even Hamlet’s death at the end, are we meant to think he has done good and will go to heaven, or has done evil and will go to hell? Who knows.

And that’s the point. If good and evil (or whatever the two sides of your spectrum) are meant to be shown as having little true meaning, you can give the player quite a lot of agency without worrying about pacing as much.

Robert McKee would probably call these character-driven stories, I suppose. It’s more about the existential angst the character faces being forced to make these decisions, going back and forth, being unsure of the path to take. These are not heroic or moralistic stories though, but storied centered on themes of change, how it can’t be avoided, but not about the benefits of how one chooses any particular strategy of dealing with it. 

What’s the point of that increased agency, if the character is being directed towards this single sort of theme? Perhaps before we can create compelling game stories that have many possible themes to explore, we have to pursue stories that have much more local agency, at the cost of conveying a smaller number of possible themes (themes here not being synonymous with actual endings, but obviously related).

I think it’s much easier, at this very moment, to create a well-paced interactive story that is perhaps somewhat depressing and fatalistic after this fashion. Why doesn’t anybody do that? Oh wait, right, it would be depressing and fatalistic.

Granted, we’re starting to see a bit more of this. Far Cry 2 seems to peruse somewhat down this train of thought, albeit forcing the player down the negative path. I haven’t gotten to it yet, because I’m in the middle of Fallout 3, which loves to present you with choices that have obvious good or evil moral implications, but no others. If I ever finish these huge ass games maybe I’ll follow up on this train of thought…

3 Responses to The low hanging mud.

  • Dave Mark says:

    Great point about the player-driven storyline. I think one of the naive assumptions that player-driven mechanics is based on is that the player also WANTS everything to work properly. Often, players are working from exactly the opposite premise: “let’s see if I can mess this up.”

    At that point, standard debugging techniques come into play. Often, developers will check the natural progression of things: “Ok… let’s do A, then B, then C, then D… Great! It works!” But what if the user/player does, C, B, D, A? You have to account for that. Simple error trapping is one thing… but constructing an algorithm with the subtle complexities of a story arc generator could be rather daunting if the player doesn’t *ahem* play along.

  • Borut says:

    Exactly – is it really feasible to try and project forward looking for some supposedly objective best arc? Or instead look back at previous player actions and try to match best what their actions are in line with, and suggest that (without actually imposing it).

  • Dave Mark says:

    Certainly we have grown a lot from this point, but I remember the early days of UO (I got in on day 2… I was on a business trip day 1.). It didn’t take long for many people to start asking “but what do you DO? What’s the plot?” At that point, MMOs were in their infancy, of course. (That would explain how their players were often infantile.)

    Eventually, as we wandered through the MMO space, people started putting together their own plots. Much effort was spent by companies in overlaying a “global storyline” to their worlds. They weren’t as rigid and explicit as campaign-based games, of course. A guide… not rails.

    Now it seems that no one really cares all that much for any sort of plot in an MMO. Sure, you can have a world definition such as that of Warcraft, Middle Earth, Conan or what not. But, aside from the artwork and the peculiarities of some of the game mechanics, does it really matter? All people want to do is show up and work the same tired formula in order to achieve whatever the near-term goal is.

    Back on point, if people are given a plot-on-rails design, they will follow it through. If they are left to their own devices, it seems that they tend to settle into some sort of mindless sediment of creative apathy.

    The hybrid of that is, as you suggested, giving them enough of a jolt to WANT to do something creative without relying explicitly on the authorship of world-engine itself. Of course, the tricky part is that you have to have give them the tools with which to do this – and have either developed the world assets to support it or be able to have the world dynamically adjust itself to adapt to what the player is doing.

    Fable 2 and the like are a step in the right direction, but may still have limitations that become obvious after a while. (Not unlike Truman discovering the walls of his sound-stage in the movie bearing his name.)