Archived

This topic is now archived and is closed to further replies.

Diodor

AI and Game Design

Recommended Posts

1) Given that the purpose of AI is to solve problems, and that of Game Design is to create them, it is clear the two intersect. Chances are problems interesting for AI research are also interesting for Game Design and can be easily imported into a game (just replace "agent" with "player"). An example: the problem of the k one-arm bandits. The agent/player is allowed a limited number of pulls on k slot machines. But each machine has a different chance of offering a gain, so the player/agent must make a trade-off between exploring slot machines he knows nothing/little about (hoping to discover a more generous slot machine) and exploiting the one he already knows offer good winnings. 2) Poor AI solutions are not necessary bad Game Design. While playing a Reversi game I'm writing, I found out I like playing against the stupid greedy AI than against the smart min-max one. The greedy solution will always try to flip the highest number of pieces, so I have an easy time knowing in advance what the computer's response will be. As a result, I always beat it (I like that), and I can aim for more challenging goals, like taking all the computer's pieces (so the game is still interesting, with different gameplay). The min-max solution just beats the hell out of me (not fun). 3) AI is generally used against the player. Using it to solve the player's problem can be a great help for game design. I'm thinking about puzzle games, the likes of Sokoban, where simple AI algorithms can pretty much solve a problem just like a person would, and offer estimations of how difficult the puzzle is, how many (if any) solutions exist, etc. Combining this with random puzzle generators, could give a lot of help for the level designer. [edited by - Diodor on August 17, 2002 5:21:52 AM]

Share this post


Link to post
Share on other sites