This paper presents a set of inference rules for reasoning about other people's knowledge. These rules are not complete, but they were effective. The paper shows how they may be used with a non-trivial set of contingent facts and rules about the knowledge possessed by various people to answer fairly complex queries about who knows what. The inference rules are based on the notion of mimicking the reasoning that might be performed by another person, rather than on an analysis of the conditions under which some fact is constrained by the other facts which may be available. The conclusion argues that the speed of the resulting system at least partially compensates for the fact that it does not manage to derive every possible conclusion.
This paper is not available online