Agents Interacting in Dialogues with Argumentation
08 / 2008 - onbekend
The world of software systems increasingly consists of autonomous artificial agents interacting in open, dynamic and unpredictable environments, such as the Internet. In such environments it is often impossible to predict all interactions the agents will have so that a centralised architecture is not feasible. Therefore, the agents' interactive behaviour must become more intelligent and flexible. For this reason, intelligent communicative interaction between artificial agents is a key research issue in artificial intelligence and agent technology. Communication languages, protocols and strategies have been studied for various types of agent interaction, such as information exchange, negotiation, and collaborative problem solving. In this field there is a particularly marked interest in the role of argumentation, i.e., the process in which agents present arguments in support of their own opinions or in attack of those of other agents. Argumentation is often claimed to promote more efficient dialogues or dialogues with better outcomes. However, so far there has been little research on whether these advantages of argumentation indeed obtain. The proposed research aims to fill this gap. The main research question therefore is: Is the use of argumentation in dialogical interaction between artificial agents beneficial for the agents or the community in which they operate? This question will be answered with systematic simulation experiments, in which forms of interaction with and without argumentation will be compared on their efficiency and quality of outcome. The research will thus contribute to a computational theory of effective communication between intelligent artificial agents.