Abstract:
A number of authors have noticed that observations of specific facts about specific states of the world cause problems for existing theories of nonmonotonic reasoning about action. We argue that these problems are pervasive in action theories, and more subtle than previously noticed. Furthermore, we note that the proposed solutions are fundamentally procedural: they involve treating observations in a separate reasoning step. These insights engender a number of open questions about the precise nature of observations in reasoning about action, and how they should be treated.