Introspection as an action in relational models

Abstract

This work studies positive and negative introspection not as ‘static’ properties an agent might or might not have, but rather as epistemic actions that change the agent's knowledge. The proposed actions include not only operations for achieving full introspection (first with respect to all formulas, and then with respect to a particular χ), but also acts for increasing the agent's introspection by only one degree. In all cases, the actions are represented as model update operations, with matching modalities for expressing the operations' effects. Sound and complete axiom systems are provided in most cases, and some properties of the operations are explored.