Multi-agent Argumentation and Dialogue

In this chapter we give an overview of multi-agent abstract argumentation and dialogue, and its application to formalise legal reasoning. The basis of multi-agent abstract argumentation is input/output argumentation studied by Baroni et al., distinguishing between the individual acceptance of agents and the collective acceptance of the system. The former may also be seen as a kind of conditional reasoning, and the latter may be seen as the reasoning of an external observer. We extend input/output argumentation in two ways. First, we introduce epistemic trust and agent communication, where the former is based on a social network representing epistemic trust, and the latter is based on so-called sub-framework semantics. Second, we introduce dialogue semantics for abstract argumentation by refining agent communication into dialogue steps. A dialogue is a sequence of steps from the framework to the extensions, where in each step of the sequence an agent can commit to accepting some arguments, or commit to hide or reveal one of his rejected arguments. The revealed arguments are then aggregated and an external observer, in our example a judge, can compute which arguments are finally acceptable at a global level.