Towards Probabilistic Argumentation

All arguments share certain key similarities: they have a goal and some support for the goal, although the form of the goal and support may vary dramatically. Human argumentation is also typically enthymematic, i.e., people produce and expect arguments that omit easily inferable information. In this chapter, we draw on the insights obtained from a decade of research to formulate requirements common to computational systems that interpret human arguments and generate their own arguments. To ground our discussion, we describe how some of these requirements are addressed by two probabilistic argumentation systems developed by the User Modeling and Natural Language (UMNL) Group at Monash University: the argument generation system nag (Nice Argument Generator) [18, 19, 20, 38, 39, 40], and the argument interpretation system bias (Bayesian Interactive Argumentation System) [7, 8, 34, 35, 36, 37].