Argumentative Deliberation for Autonomous Agents

This paper presents an argumentation based framework, developed as an extension of an existing framework for nonmonotonic reasoning, in order to support an agent’s self deliberation process. The framework allows the agent to draw conclusions taking into account in a natural way a given preference policy. After developing the argumentation framework we examine two general cases of such argumentative deliberation: (a) under a preference policy that takes into account the roles agents can have within a context pertaining to an environment of interaction and (b) under a preference policy for the current needs of the agent emerging from his profile. In the first case we apply the argumentative deliberation model within a simple agent interaction scenario where each agent’s selfdeliberation determines, according to his own policy, his position at each step of the interaction process. In the second case we apply the framework to model motivational factors that apparently drive human behaviors and therefore can define agent personalities. Agents will thus similarly, as it is claimed in psychological literature for human beings, chose at any moment to pursue, those goals that are most compatible with their current motivations. The proposed argumentation framework allows us to define policy preferences at different levels of deliberation resulting in modular representations of the agent’s knowledge or personality profile. This high degree of modularity gives a simple computational model in which the agent’s deliberation can be naturally implemented.