On Generating Abstract Explanations via Knowledge Forgetting

Stylianos Loukas Vasileiou, William Yeoh

Abstract: In this paper, we investigate the problem of generating explanations from the context of Human-aware AI Planning. Particularly, we focus on an explanatory setting for tasks encoded in a logical formalism, where given an agent model (encoding the task), an explanandum entailed by the agent, and a user vocabulary specifying terms in the task, the goal is to find an explanation that is at an appropriate abstraction level with respect to the user's vocabulary. We propose a logic-based framework aimed at generating such explanations by leveraging a method called \emph{knowledge forgetting}, and present an algorithmic approach for computing them. Our experimental evaluation shows the promise of our framework.