Blissful Ignorance: Knowing Just Enough to Coordinate Well

To coordinate, intelligent agents might need to know something about each other, and about themselves, and about how others view themselves, and how they view others, and how they think others view others, and so on. Taken to an extreme, the amount of knowledge an agent might possess to help make sense of an interaction might outstrip the agent’s limited reasoning capacity (its available time, memory, etc.). Much of the work in studying and building multiagent systems has thus been devoted to developing practical techniques for achieving coordination, typically by limiting the knowledge available to or necessary for agents---that is, by keeping agents selectively ignorant. In this paper, I try to put various such efforts in perspective, characterizing different ways in which agents can beneficially be kept ignorant of some knowledge. *