Arguing On Communicative Goals In Explanations

This paper describes how we simulate a dialog aimed at resolving conflicts between an Instructing and an Executing Agent in explanation text planning. We describe how the two agents are modelled. We then illustrate the algorithm which, starting from the discovery of a conflict and proceeding through a set of negotiated belief revisions, brings to solve the conflict or to conclude that it cannot be solved. An example in the domain of drug prescription is provided. 1. INTRODUCTION In this paper we consider a particular case of collaborative activity, where an Instructing Agent (IA) has to explain to an Executing Agent (EA) how to perform an action. However, IA's and EA's views about whether to talk about a specific subject can be in conflict: IA might consider irrelevant or inappropriate to talk about subjects which, on the contrary, EA finds interesting, or the inverse. To solve this conflict, the two agents should be able to initiate a dialogue aimed at understanding the reasons of t..