Editor's Choice

Editor's Choice
Next-Generation Storytelling

Warren Spector | 24 Apr 2007 08:00
Editor's Choice - RSS 2.0

Not long ago, I had a conversation with Doug Church, secret master of gaming, where he said something like this: A story is constructed of sentences, strung together in a coherent, dramatically significant order. Game "sentences" are the actions available to and selected by a player. The more sentences we allow players to construct (in other words, the deeper the pool of options we offer), the cooler and more numerous the story possibilities will be. To that extent, a robust world and character simulation - both made possible by next-gen hardware - will allow us to tell a better story. But there's a hitch: all the graphics power of the new platforms.

We've made - and, thanks to the new hardware, will continue to make - great strides in the fidelity with which we can portray a world. Our characters will look even better. Our worlds will look and feel much more convincing than they ever have. And audiences will come to expect a certain level of believability in the worlds they explore. They will expect the world to look and behave the way the real world does. ("It looks real; it'll act real.")

All of that means AI - cornerstone of creating great characters and, therefore critical to great story games - becomes even more challenging. And here I'm just talking about the fundamentals of navigation and base level interaction. We've made great strides in AI over the last few years, but you'd hardly know it - the advances have come in the service of "just keeping up" with graphical and simulation enhancement.

Back in the day (that is 20, 10, maybe even five years ago), NPC's just had to navigate through a 2-D world or a simple 3-D one. Now, even in a relatively simple game, they have to deal with highly complex 3-D spaces.

Again, Doug made the point the other day: "We didn't used to have to worry," he said, "about a glass full of water getting knocked over, wetting the pants of a character seated at the table and then having to deal with the NPC's response, other characters' responses or the player's potential responses." Now, or soon, players will expect the AI to react believably to all of that.

Once we can create beautiful, photorealistic spaces, players will expect us to do so, and then we'll have to teach our NPCs how to navigate through and interact with ever-more complex worlds appropriately. Rapid advancements in graphical fidelity and depth of simulation have always left AI and design playing catch-up. That problem seems likely to get worse, not better, in the next few years.

Most terrifyingly, we may, in fact, have missed our opportunity to play with story-friendly AI and procedural story generation. That moment may have come and gone back in the day of lower-fidelity world sims. Now that we're in a world of high-fidelity worlds, all of our energy is likely to be sucked up just maintaining the levels of AI we already have! The cause of non-navigational/non-combat AI will take a backseat again.

In the short term, at least, next-gen hardware will allow us to make prettier games (at great cost) but not better games, and there's nothing inherent in the hardware that makes better stories more or less likely.

What Do We Do About It?
Those of you who know me - or know my games - know I can't have just one endgame. That's too limiting for players.

Comments on