We use the term information state to mean, roughly, the information stored internally by an agent, in this case a dialogue system. A dialogue move engine updates the information state on the basis of observed dialogue moves and selects appropriate moves to be performed. In this paper we use a formal representation of dialogue information states that has been developed in the TRINDI, SDS and INDI projects*.
The structure of this paper is as follows: First, we give a brief description of a general dialogue system architecture which can be used for experimenting with different kinds of information states and dialogue move engines. We explain the distinction between static and dynamic information state, and discuss how rules formulated in terms of conditions and operations on information states can be used to (1) update information states based on observed dialogue moves and (2) select dialogue moves based on the current information state. We present a particular notion of dynamic information state based on Ginzburg's theory of Questions Under Discussion (QUD) [Ginzburg1996a,Ginzburg1996b,Ginzburg1998]. An experimental dialogue system which uses this notion of information state is presented. We then look at the role of accommodation in information state transitions and point to examples of two kinds of accommodation: accommodation of questions under discussion and of dialogue plan. Finally, we argue that accommodation should be associated with update rules, or tacit moves (not associated with an utterance), and show how the implementation of these rules yields improved behaviour in the experimental dialogue system.