Reasoning about Incomplete Agents
Hans Chalupsky, Marina del Rey, Stuart C. Shapiro · 2000
We show how the subjective and nonmonotonic belief logic SL formalizes an agent’s reasoning about the beliefs of incomplete agents. SL provides the logical foundation of SIMBA, an implemented belief reasoning system which constitutes part of an artificial cognitive agent called Cassie. The emphasis of SIMBA is on belief ascription, i.e., on governing Cassie’s reasoning about the beliefs of other agents. The belief reasoning paradigm employed by SIMBA is simulative reasoning. Our goal is to enable Cassie to communicate with real agents who (1) do not believe all consequences of their primitive or base beliefs, (2) might hold beliefs different from what Cassie views them to be, and (3) might even hold inconsistent beliefs. SL provides a solution to the first two problems and lays the groundwork to a solution for the third, however, in this paper we will focus only on how agent incompleteness can be handled by integrating a belief logic with a default reasoning mechanism. One possible application of SL and SIMBA lies in the area of user modeling. For example, Cassie could be in the role of an instructor who, among other things, has to deal with the incomplete beliefs of her students.