Ghosts In The Machine

The November, 2006, issue of the Atlantic has a feature about videogames called Sex, Lies, And Videogames, written by Jonathan Rauch. (What’s with the thoughtful magazines coming out with videogame pieces? Where are you, New Yorker?)

The article is an interesting take on the attempt to broaden the appeal of games, mainly by improving characters. Most of the article is taken up with a discussion of Façade, a conversation-based game, or what the article calls “interactive drama.” As a version of the “are games art?” discussion, this one has a lot more reporting and details than most, which is nice.

However, the big problem of games like Façade is the AI, which, right now, promises much more than it delivers. And I’m skeptical it will reach the levels necessary to make a good game anytime soon. The article is not available to non-subscribers, so I’ll pull some choice bits for examination after the jump.

Here’s the main thrust of the article:

In certain rarefied circles of AI academia and video-game design, people sometimes theorize about a computer program that would combine the graphical realism of a modern video game with the emotional impact of great art. “Interactive drama,” the concept is called. It might contain artificial people you could converse with, get to know, and love or hate. It might engineer dramatic situations, complete with revelations and reversals. Entering this world, you would feel as if you had been thrust into the midst of a soap opera or a reality-TV show.

Apart from a minor quibble—I think a number of games have already achieved the emotional impact of great art—I think this is a fine goal.

As Andrew Stern, one of the Façade designers, says in the story, “There’s no drama genre, there’s no comedy genre … What exists right now are action movies, basically.” More comedy, more drama, nothing necessarily wrong with that. Rauch continues:

If videogames seem inhuman, that is because they lack humans. Their esoteric syntax is an artifact of a stunted environment in which blasting someone’s head off is easy but talking to him is impossible.

I have to admit: I’m immediately struck by a defensive impulse. Why do we need deep drama from games? The background assumption seems to be that games are limited. Why can’t we let them be as they are and do what they do best? People don’t bemoan the lack of comedy in sculpture. Are we trying to shoehorn things into a medium in which they don’t fit? I’m inclined to side with people like Steven Johnson, who have argued that games can’t and shouldn’t be judged with the same criteria that we judge, say, novels.

But I recognize games often have humans undertaking actions in a world filled with other humans, and in that case we can link it with drama. And I want to see games evolve to be as broad, popular, and varied as possible, so let’s explore all possible game designs at this early stage in game history. If some games had realistic characters that responded to players, that would be swell. I’m all for “redefining the meaning of video-game ‘play,’” as the story puts it. (Although I still want to be able to race cars and blow heads off.)

I know more non-gamers would get into videogames if they could talk to the characters. I’ve been playing Façade over the last few days, and my wife snatched the computer away from me. Talking with characters is just what she wants in a game, and the simple interactions of Animal Crossing and Knights of the Old Republic only go so far with her.

The problem is, the conversations in Façade are very poor. When players aren’t being led by the nose by the characters of Façade asking yes or no questions, at best it seems like the characters in the game pick up on mood words like “angry” or exciting terms like “sex” to steer the conversation. Often, the meaning of conversations feels like a result of pattern seeking on the player’s part.

It seems to boil down to a problem of AI. If you’ve ever talked to a chat bot online, you’ll know what I mean. The most successful ones seem to be those that give ambiguous answers that you can imagine were related to your statement. And talks I have had with people who work in AI tend to be pessimistic, almost suggesting that even seeking a virtual intelligence that you can talk to is a silly task, and the goal of passing the Turing test an out-dated fantasy.

But I’m out of my depth here. Maybe I’ve just heard from the pessimists. Maybe there are revolutions just around the corner. Michael Mateas, another designer of Façade, predicts it will be “totally doable within twenty years.”

Ultimately, I’m a wishful pessimist when it comes to believable AI in videogames. As someone from Electronic Arts tells Rauch in the article:

A lot of people have worked on it. Every year we’re like, “We’re going to design incredibly intelligent, fluid humans who act realistically.” We try to take this huge step—and we fall all the way back down. At least [Mateas and Stern] ended up somewhere new. It doesn’t all work, but at least it is a step.