An agent operating in the real world must often choose from among alternatives in incomplete information environments, and frequently it can obtain additional information about them. Obtaining information can result in a better decision, but the agent may incur expenses for obtaining each unit of information. The problem of finding an optimal strategy for obtaining information appears in many domains. For example, in ecommerce when choosing a seller, and in solving programming problems when choosing heuristics. We focus on cases where the agent has to decide in advance on how much information to obtain about each alternative. In addition, each unit of information about an alternative gives the agent only partial information about the alternative, and the range of each information unit is continues. We first formalize the problem of deciding how many information units to obtain about each alternative, and we specify the expected utility function of the agent, given a combination of information units. This function should be maximized by choosing the optimal number of information units. We proceed by suggesting methods for finding the optimal allocation of information units between the different alternatives.
展开▼
