Using Uncertainties as Basis for Evaluating Plans

摘要

Decision-making requires coping with uncertainties. Technical decision-support has traditionally focused on trying to represent these uncertainties and perform reasoning under uncertain conditions using probability theory or one of its relatives. An alternative method for handling uncertainty is to study what decision-makers actually do when encountering uncertainty and try to develop a system that supports or utilizes those activities. This paper suggests an agent based system that gives the decision-maker an opportunity to evaluate what impact different assumptions have on a plan. The system tracks uncertainties encountered during planning and use these uncertainties to generate a set of scenarios. The scenarios are then tested for workability in a cooperative process between the planners and the system. The purpose of such evaluation is to make sure the plan is workable in a worst-case scenario and to identify the uncertainty(s) that has most impact on the mission.