Personalized information dissemination using agents organizations

The multi-agent system described in this paper implements a collaborative filtering process that applies personalization techniques to measure the appropriateness of a document for a community of users. Collaborations among agents and users are defined in a workflow that describes the broadcasting of documents, which flow through a series of filters at different levels (community agents, personal agents, and users). With this system we demonstrate that a workflow definition (roles, tasks, and their relationships) is a key component for the specification of the multi-agent system organization, improves the design of agents interactions and coordination, and supports the identification of agents' goals and tasks. Furthermore, we discuss social rules that can be used to control the global behavior of the community of agents.