The theme of Corvus Elrod’s Round Table for this month (which I may as a tendency take as a general monthly theme) is the role of artificial intelligence in gaming. In this context, as I understand it, AI most often refers to intelligent agents specifically; that is, the decision-making processes behind enemies, allies, social NPCs, and any other game objects meant to project an air of sentience.
I’ve had some small professional experience related to this particular field. The major recurring theme I recall from my study of the disposition of these intelligent agents is this: their human counterparts are, on the one hand, enthusiastic beyond measure to attribute motivation to every random or accidental behavior the agent displays and, on the other hand, consistently unimpressed by any and all intended behaviors.
While frustrating to the designer, this is interesting in what it suggests about how people react to a program attempting to appear intelligent. We are eager to attribute sentience to that which we don’t understand, yet quick to reject this hypothesis when an underlying mechanism peeks through. This makes some sense from an evolutionary standpoint; it is probably safest to assume that anything you don’t understand may potentially wish you ill. But the more knowledge we have about the world, the less we see spirits in it.
The simplest implication for videogames is that greater randomness in the algorithm may contribute to a greater impression of sentience – as in, say, The Sims 2, where conflicting drives and constant random rolls lead to a relatively unpredictable, and vivid, stable of NPCs. Their personalities (such as they are) are just arbitrary numbers, but that may be why they make such an easy target for the feverish imaginations of the players.
But there’s more to it than this, of course. We are equally eager to attribute sentience to characters whose motivations are drawn in a traditionally literary fashion, bypassing algorithms, such as those in Knights of the Old Republic. But this seems intuitively to be a different kind of sentience, and I’m not sure whether it carries over once your exhaustively detailed conversation ends and the battle begins. What is the relationship between AI and character? I’ll leave that question for another night.
For more posts on this topic, please visit the Gaming Blog of the Round Table home page.
4 responses so far ↓
Corvus // July 19, 2007 at 4:42 am
Great inaugural post, Caroline! Since you’re on WordPress and the IFRAME script won’t work, can I get you to include a link to the Round Table’s home page at the bottom of the post? Something along the lines of:
For more posts on this topic, please visit the Gaming Blog of the Round Table home page.
Then you can delete this comment if you like. Thanks again!
Chris // July 25, 2007 at 12:08 pm
The question this sparks in me is: If most of what AI achieves occurs entirely in the player’s head, why bother researching AI…? >:)
Caroline // July 26, 2007 at 3:05 pm
Thanks for the response, Bateman! Well, the short answer is, if perception occurs in the player’s head, why bother researching graphics? The point of both graphics and AI (in the sense of intelligent agents) as I see it is to encourage and manipulate those internal processes of perception and social cognition as much as possible.
And just as graphics was historically limited by a focus on physical photorealism, I think that the study of intelligent agents is probably limited by the lingering influence of strong AI, as you suggest in your post. If a non-photorealistically rendered (NPR) game like Okami can make such a strong visual impact on gamers, it may make sense to start exploring cognitively unrealistic AI.
As for what that means, I’m not sure. But if researchers in NPR have taken their inspiration from the visual arts, maybe cognitively unrealistic AI should explore literary and theatrical forms of character building? I’m imagining agents trained to respond to stimuli like soap opera characters, here. And I like it.
Marcus Riedner // August 1, 2007 at 4:00 pm
Discussions on AI always make me want to slide into quasi-mysticism. Much like the players you are describing here, I am forever finding coincidence and chance playing a role in how I interpret the actions of AI agents.
At the end of the day, I have yet to see an AI agent in action that really meets my needs as a player. That need, of course, is companionship and a deeper connection to another being. Particularly in MMO’s I am regularly missing a deeper connection, as I watch friends and strangers save the same princess who’s been poisoned over and over. Or watch the mindless spiders walk back and forth along a path.
I think the people react as they do to programs trying to act intelligent because, well, they either act too intelligent, or they are not intelligent enough. At least when an agent does something unexpected and random it gives the player room to try and reason out that action…