Managing information flow for complex, dynamic tasks using multi-agent collaboration

In this paper, we present a multi-agent information system model that can deal with complex multi-activity information-driven tasks. After defining a single information agent architecture in an abstract way, we describe the multi-agent model (called the parallel model), as well as a simple collaboration protocol for the multiple activity information agents. Experimental results are presented with the multi-agent system that demonstrate that a significantly larger number of relevant documents can be retrieved through collaboration than what is possible with decoupled agents. Hence, such multi-agent parallel models appear to hold considerable promise in providing the user with high quality information services in a variety of business and scientific information-driven tasks.