Progressive Reasoning for Complex Dialogues among Agents

Abstract

In this paper we propose an architecture for deliberative agents based on progressive reasoning. When an agent receives a query, it tries to satisfy it by building an answer based on its current knowledge. Depending on the available time or the urgency of the requirement the agent can produce answers with different levels of quality. Agents could build progressively their answers with the information they receive from perception or during the dialogue with other agents. We assume that in the real world normally it is better to receive an answer with poor quality than no answer. The answer can be good enough for the receiver or the receiver can spend more time to wait for a better answer. Autonomy implies taking the best decision with the available information, avoiding blocking situations and no action.