Explaining the outcome of knowledge-based systems: A discussion-based approach

Many inferences made in everyday life are only valid in the absence of explicit counter information. This has led to the development of nonmonotonic logics. The kind of reasoning performed by these logics can be difficult to explain to the average end-user of a knowledge based system that implements them. Although the system can still give advice, it is hard for the user to assess the rationale behind this advice. In this paper we propose an argumentation approach that enables the advice to be assessed through an interactive dialogue with the system much like the discussion one might have with a colleague. The aim of thie dialogue is for the system to convince the user that the advice is well-founded.