CAISOR Archive of Articles, PM-2007-002

Reification of Action Instances in the Leonardo Calculus

Erik Sandewall

Proceedings of IJCAI Workshop on Reasoning about Actions and Change, 2007.

Open access: [pdf] [ps]
Publisher version: [No link from here].

This article describes the Leonardo Calculus, a logicist representation for actions and change where action instances are reified. This makes it possible to represent relationships between actions, such as the relation between an action and its subactions in a hierarchical action structure. It also makes it possible to represent the relation between an action and the objects that have been created in it, for example, the relation between the action of building a house and the house itself, or the relation between the action of understanding a sentence and the parse tree that was an intermediate representation for the sentence.

Since action instances are reified, the logic must limit the set of action instances in each model to those that are actually warranted by known knowledge, which suggests minimizing the domain of action instances, rather than one or more predicates. However, somewhat surprisingly, this calculus ends up minimizing a predicate anyway.