EASSS-07: Argumentation in Agent Systems
This is a WWW-page with resources related to the EASSS-07 tutorial of Martin Caminada and Henry Prakken on Argumentation in Agent Systems.
In order for agents to be able to reason with uncertain information or "rules of thumb", some model of defeasible reasoning is required. The disadvantage of many of today's formalisms for defeasible reasoning is that they have a mathematical nature that is quite remote from how people actually perceive their everyday commonsense reasoning. This makes it quite difficult for people to understand and trust the behavior of an agent system. The argumentation approach is to bridge this gap by coming up with a formalism that is rigid enough to be implemented in an agent system, while at the same time close enough to informal reasoning to be understood by users.
In the current course we discuss the fundamental concepts and structure of argumentation formalisms. We also show how existing approaches to defeasible reasoning (such as default logic) can be seen as instances of argumentation. Furthermore, we discuss how some instances of argumentation can be specified in the form of a dialectical process in which a proponent and an opponent discuss about the validity of a main claim. This enables agent designers to specify a communication protocol that adheres to desirable logical properties. We also discuss some problems and pitfalls that are associated with today's generation of argumentation formalisms, as well as how to deal with this. In short, the course will give an overview of the things that every agent designer should know before enhancing his agent system with defeasible reasoning capabilities.
The tutorial slides are available at:
Material from the following articles is discussed in some detail during the tutorial.