The Frame and Focus Problems in AI: Discussion in Relation to the Believer System

There are two types of reasoning about actions - Planning and Interpretation. We are involved in the Interpretation of observed sequences of actions using a psychological theory of act interpretation. The Interpreter uses a model space to maintain a current description of the observational world. There are two kinds of problems in maintaining a model - Consistency with respect to what we know about the domain and Correctness with respect to an external observable world. Past discussions of the problem have dealt mainly with the Issue of consistency. The issue of ensuring correctness is a difficult one and this paper proposes the Heuristic of using the detected inconsistencies of the model state as a guide to selectively removing disparities between the model state and the observable world. A Knowledge Representation Framework is described which allows all aspects of the Frame Problem to be tackled coherently. An example is given to illustrate our approach and an indication of the rule learning capability of the system Is given. The Focus problem for the Observer is introduced and Is shown to be complementary to the Frame Problem for the modelling system.