Questions Under Discussion and Dialogue Moves

Broadly, the topic of this paper is the relation between agents' information states and dialog moves (i.e. questions and answers). It's about information goals, and planning of how to get them. It explains some simple pragmatic phenomena: the question "Do you know the time?" indicates a desire to know the time, which is how listeners know that a literal answer is not desired.

The paper is about comparing and combining the QUD model (basically, a stack of questions to be discussed, i.e. goals and sub-goals) with the "classical" BDI (which, in general, might need full-blown epistemic logic (to focus only on the "B" from "BDI")). Despite their discussion at the end of section 4, my inclination is to just adapt BDI (the more expressive formalism) to deal with those issues. But I'll readily admit that I'm not familiar with these formalisms.

As with all papers with logic content, I wish this came with an applet to allow me to play some of these formal dialog games.

It's a very neat sort of thing, because it formalizes common-sense. I can even imagine this applied to tutoring social skills to autistic people.