Reasoning about and Discussing Preferences between Arguments

Abstract

Agents that have different knowledge bases and preferences over arguments can use dialogues to exchange information and explanations. In order for the dialogue to be useful, agents need to utilize the other participants’ knowledge fully while being resistant against manipulation. Furthermore, the information they exchange can be objective but also subjective such as what goals an agent wants to achieve. To understand why another agent draws a certain conclusion it is necessary to understand and communicate preferences over arguments. This paper proposes an ASPIC-based meta-level argumentation logic for reasoning about preferences over arguments. Extended argumentation frameworks are used to determine what arguments are justified. Prakken’s dialogue framework is then adapted for meta-level arguments and a protocol is proposed that explicitly distinguishes between objective and subjective topics. Several mechanisms for using other agents’ knowledge have been proposed in the literature. This paper proposes to use different acceptance attitudes with respect to claims made in a dialogue and to store the source of those claims on a meta-level. In the meta-level, agents can then reason about the effect of other agents’ claims on the conclusive force of arguments. This makes agents more robust against manipulation and able to handle new information better.

The research reported here is part of the Interactive Collaborative Information Systems (ICIS) project, supported by the Dutch Ministry of Economic Affairs, grant nr: BSIK03024.