Sign up for news

Lost my wallet this weekend (I prefer to think of it this way over the more insidious option, that it was stolen). No cash, can’t go out to get breakfast (and I prefer this term over the less manly brunch). So I figure, Hey, let’s go for a bike ride! Flat tire. The world does not want me to leave the goddamn apartment.

Thankfully, Andrew has provided some morning reading material. Lots of good stuff, but what I’ll write about in particular today is gonna be Ian Horswill’s position paper on modeling psychopathology in AI. He argues that we should have more AI models for with characters that have the same failings/weaknesses in other narrative mediums.

Now, I’m a fan of this line of thinking fer sure (and the general thought that we should be trying to simulate characters, and not people). But, this paper only argues for half of that – what about the modeling of character strengths? To make a character compelling they can’t be a total loser (all weaknesses), or perfect (all strengths).

Most interesting to me, though, are those characters whose strengths also serve as weaknesses, but in different situations. Take Adama from Battlestar Galactica (the new one, naturally). He’s incredibly loyal, but but his loyalty to Tigh gets him in trouble because of all of Tigh’s problems & bad decision making. You see the characteristic in times of strength, and then you understand the character better when it becomes a weakness – in other words, they can do something incredibly stupid and that might actually increase your connection to them. While if they did something stupid and you didn’t understand why, your interest in the story would drop dramatically.

Wow, so you mean a character can do something dumb and you *enjoy* this? Compare that to most AI in game’s today – generally speaking, that’s hard to find. It is a problem you come across, though, even the most basic button mashing brawler. Take, for instance, the mad goblin bomber. We’ve seen him in any number of fantasy games. You know his deal, he runs at you with a bomb, and explodes.

Now, strictly speaking, this is about the stupidest thing any character could conceivably do, in terms of trying to defeat someone who has slain hundreds of their kinsmen. Yet I’ve never once heard anyone say, “Goddamn, that’s some stupid AI!” He runs right at you and explodes! Into little bitty bits! C’mon, people!

But we forgive this – why? Because he wildly gesticulates his hands in the air, screams unintelligibly, and probably dresses in wild colors compared to his saner goblin friends. We understand his character (as such), is not prioritizing a rational attempt to deal with you as a threat in a strategic fashion. This is a character that is fascinated with blowing shit up. This is often a strength, mind you; in this case, not so much. But we can understand his priorities in decision making, even if they are weaknesses (as they need to be for the player to complete the game) and so we understand his character, which in turns makes the AI not look stupid.

Slowly, ever so slowly, this will hopefully evolve past the mad goblin, as Horswill suggests.