Tech Reports

ULCS-04-010

Identifying Audience Preferences in Legal and Social Domains

Paul E. Dunne and Trevor Bench-Capon


Abstract

Reasoning in legal and social domains appears not to be well dealt with by deductive approaches. This is because such reasoning is open-endedly defeasible, and because the various argument schemes used in these domains are often hard to construe as deductive arguments. In consequence, argumentation frameworks have proved increasingly popular for modelling disputes in such domains. In order to capture a third phenomenon of these domains, however, namely that rational disagreement is possible due to a difference in the social values of the disputants, these frameworks have been extended to relate the strengths of arguments in the dispute to the social values promoted by their acceptance. If we are to use such frameworks in computer systems we must be aware of their computational properties. While we can establish efficiently the status of an argument for a particular audience, deciding the status of an argument with respect to all potential audiences is known to be intractable. The main result of this paper is an algorithm which identifies the audiences for which some set of arguments is coherently cotenable. The usefulness of this algorithm is particularly evident in the dialectical settings in which these value based frameworks are most naturally deployed.

[Full Paper]