The LCC language of agent dialogue (at least the version I worked
with during my MSc) does not always permit dialogues to complete
correctly. If constraints are broken, the dialogue simply stops.
In
this talk I will address this problem by offering a formal method to
overcome constraint satisfaction failure by searching for other
possible dialogue paths. This is achieved by 'inducing backtracking'.
I'll also be discussing the problems faced and the limitations of this
method.