Reasoning about goals to resolve conflicts

A mechanism is presented for negotiating to resolve conflicts that arise when autonomous agents need to agree on a collection of operations in order to achieve a goal. These conflicts arise when the agents have different policies as to which operations are acceptable. The authors' mechanism protects the privacy of the agents because they are not required to exchange information freely about their goals. Instead, each agent can infer the goals of the other agents. This characteristic is obtained by incorporating plan recognition in the negotiation process. Their mechanism requires agents to find alternative activities that achieve the goals of all involved agents, rather than allowing them to rely on modification of their own or other agents' goals. This approach is reasonable in contexts such as telecommunications systems, where agreement about end goals is usual, and the difficulty is in achieving agreement on the means to achieve the end goals. The negotiation mechanism is illustrated using an example where agents represent subscribers to a telecommunications system.<>