Artificial Intelligence

13

The Propositional Calculus

13.1 Using Constraints on Feature Values

I have discussed two rather different methods for modeling an agent’s world: iconic based and feature based. Binary-valued features are descriptions of the world—what is true about it and what is not true about it. Iconic representations are simulations of certain aspects of the world. Although simulations are more direct and often therefore more efficient than descriptions, descriptions have a number of advantages of their own. The values of features are easy to communicate to other agents, whereas iconic models may not break up well into isolated pieces for incremental communication. Computing the values of features often requires less elaborate perceptual processing than would ...