A model of structured discourse for cooperating intelligent agents

Resolution of large problems in an organizational context often requires decomposition of the problem into subproblems which will allow individual agents to work on the subproblem of their expertise and resolve the overall problem through coordination with each other. Examples of such problems include engineering and software design, project planning, product design and marketing, and auditing decisions. We address the issue of designing information sharing and cooperation mechanisms to support intelligent agents in the overall problem resolution process using distributed AI techniques. The central mechanism that renders the subproblems to be relatively independent is the truth maintenance system at the individual agent (intra-agent) level which allows a consistent integration of observations, assumptions, and rules to arrive at propositions regarding a subproblem and allows the member to maintain a non-conflicting set of beliefs. We propose a discourse protocol which allows agents to communicate in a structured fashion and an inter-agent truth maintenance system which allows multiple agents to identify conflicting beliefs. A conflict resolution strategy allows agents to resolve both internal conflicts and most inter-agent conflicts. We illustrate the working of the model and discuss the design of a prototype system