FORMALIZING NARRATIVES USING NESTED CIRCUMSCRIPTION

Representing and reasoning about narratives together with the ability to do hypothetical reasoning is important for agents in a dynamic worl d. These agents need to record their observations and action execution s as a narrative and at the same time, to achieve their goals against a changing environment, they need to make plans (or re-plan) from the current situation. The early action formalisms did one or the other. F or example, while the original situation calculus was meant for hypoth etical reasoning and planning, the event calculus was more appropriate for narratives. Recently, there have been some attempts at developing formalisms that do both. Independently, there has also been a lot of recent research in reasoning about actions using circumscription. Of p articular interest to us is the research on using high-level languages and their logical representation using nested abnormality theories (N ATs)-a form of circumscription with blocks that make knowledge represe ntation modular. Starting from theories in the high-level language L, which is extended to allow concurrent actions, we define a translation to NATs that preserves both narrative and hypothetical reasoning. We initially use the high level language L, and then extend it to allow c oncurrent actions. In the process, we study several knowledge represen tation issues such as filtering, and restricted monotonicity with resp ect to NATs. Finally, we compare our formalization with other approach es, and discuss how our use of NATs makes it easier to incorporate oth er features of action theories, such as constraints, to our formalizat ion. (C) 1998 Elsevier Science B.V. All rights reserved.