Argumentation has proven to be a useful abstraction mechanism for understanding several problems, for example non-monotonic, defeasible reasoning in artificial intelligence, legal reasoning, several forms of practical reasoning performed by intelligent agents, medical decision-making, and security.

In order to provide tools to solve these problems, several computational frameworks for argumentation have been proposed, often based upon Dung's abstract argumentation. This form of argumentation focuses on determining the "acceptability'' of arguments based upon their capability to counter-attack all arguments attacking them. In abstract argumentation these arguments and attack relation between arguments are seen as primitive notions, defined entirely abstractly, and this allows for intuitive and simple computational models, but does
not show how to find arguments and how to exploit the fact that different arguments are built from the same premises. Assumption-based argumentation is a general-purpose framework for argumentation, whereby arguments and attack relation are not primitive concepts, but are defined instead in terms of deductions from assumptions and contraries of assumptions.

In this talk I will describe assumption-based argumentation, how it relates to abstract argumentation, several computational models for assumption-based argumentation, a family of systems implementing these models and some applications. I will also discuss the limitations of these systems when deployed by non-expert users.

Watch the webcast replay >>