A characterization of types of support between structured arguments and their relationship with support in abstract argumentation

ANDREA COHEN; SIMON PARSONS; ELIZABETH I. SKLAR; PETER MCBURNEY

INTERNATIONAL JOURNAL OF APPROXIMATE REASONING

ELSEVIER SCIENCE INC

Año: 2018 vol. 94 p. 76 - 104

0888-613X

Argumentation is an important approach in artificial intelligence and multiagent systems, providing a basis for single agents to make rational decisions, and for groups of agents to reach agreements, as well as a mechanism to underpin a wide range of agent interactions. In such work, a crucial role is played by the notion of attack between arguments, and the notion of attack is well-studied. There is, for example, a range of different approaches to identifying which of a set of arguments should be accepted given the attacks between them. Less well studied is the notion of support between arguments, yet the idea that one argument may support another is very intuitive and seems particularly relevant in the area of decision-making where decision options may have multiple arguments for and against them. In the last decade, the study of support in argumentation has regained attention among researchers, but most approaches address support in the context of abstract argumentation where the elements from which arguments are composed are ignored. In contrast, this paper studies the notion of support between arguments in the context of structured argumentation systems where the elements from which arguments are composed play a crucial role. Different forms of support are presented, each of which takes into account the structure of arguments; and the relationships between these forms of support are studied. Then, the paper investigates whether there is a correspondence between the structured and abstract forms of support, and determines whether the abstract formalisms may be instantiated using concrete forms of support in terms of structured arguments. The conclusion is that support in structured argumentation does not mesh well with support in abstract argumentation, and this suggests that more work is required to develop forms of support in abstract argumentation that model what happens in structured argumentation.