Reaching informed agreement in multispecialist cooperation

Very often, complex decisions must be made by a group of specialists rather than a single decision maker. To make an effective decision, the combination of the group's expertise must be brought to bear on the situation. Fusing expertise where individuals have very detailed knowledge in their own areas and much weaker understanding of others is characterized by many difficulties: (1) agents cannot communicate their expertise in an intelligible way to nonexperts because of differences in vocabulary and conceptual content; (2) the process allows for incorrect inferences; and (3) no one knows what anyone else needs to know. This impasse cannot be broken until shared mental models are developed to provide a level of agreement in evaluating alternatives needed to focus the activity of the group. This article presents a model of decision making by teams of specialists in which agents' evaluations confound expert and naive inferences in judging alternatives. A partitioning of agent knowledge into expert and naive models is proposed. The naive portion of agents' models provides both a common language and the inferential skeleton needed for the development of shared models. Communications are categorized into types of evaluation or justification based on their form and the entities they involve within the agent models. A process of model refinement is outlined, linking communications among agents to modifications of the naive/ shared portions of their models. The process of cooperative problem solving by a team of specialists is characterized as a search among alternatives in which model refinement continually alters the agents' evaluations, leading to progressively greater accuracy and more precisely directed search. The model is intended as a research tool for investigating multiagent problem solving among people and machines.