ABSTRACT
This chapter introduces the idea of maintaining probability and possibility distributions over some unobserved elements of game state rather than simulating their evolution explicitly. If such distributions are used correctly, they allow the artificial intelligence to defer decisions relating to unobserved elements of game state without the player realizing that anything unusual is happening. A probability map provides a mapping from game states to probabilities that represent the relative likelihoods of the game being in different states. For example, the location of a character in a village is part of the game state and can be approximated by discrete locations such as the character’s home, the stables, the tavern, the well, the shop, and the stretches of road between them. For an application of probability and possibility maps to be convincing, it is important that they are updated in a way that is consistent with the player’s observations of the game’s state.
