Modelling Argumentation with Belief Networks

We present a probabilistic framework in the logic of counterarguments which may be useful to create a computational model of arguments in natural language. An appropriate knowledge representation is a good starting point for simplifying the solution of AI problems. In this paper, we advocate for the use of belief networks as a formalism for representing ar- gumentation exchanges. Although we will now consider only their use in simulating the reasoning process (evaluation of argumentation strategies and recover after undermining attempts), it is our future plan to show also their properties in favoring automatic generation of messages. We start from some example sentences in the healthy eating domain, to analyze them in the light of Walton's argumentation schemas (15): 1. "The FDA Agency in the UK, which is an expert in healthful living, says that eating vegetables is a form of healthy eating: so, this should be taken as true. "