Previous research has suggested that conveying information about uncertainty in linguistic form does not necessarily lead to decisions that are inferior to those made on the basis of numerical probability estimates. This research will build upon the earlier work by studying how a decision maker (DM) combines and uses vague and precise information about uncertainty. Vague information generally arises from sparse, indirect, or unreliable observations, and often is communicated linguistically (for example, in a forecast by an intelligence analyst that a particular event is likely to happen). Precise information is generally the result of direct relative frequency data, and often is communicated numerically (for example, in predicting failure rates of equipment parts on the basis of historical data). This research will develop two new realistic decision paradigms that allow control over the nature and degree of information available to the DM. In all cases, DMs will choose among or bid for lotteries involving outcomes contingent on the occurrence of uncertain events. In one paradigm the events will consist of factual claims being true (rather than false). The nature of the DMs' uncertainty about the claims will be manipulated by selecting suitable factual domains. In the other paradigm the events will consist of objects moving along tracks on a computer screen and reaching targets (rather than being stopped by a barrier on the track). Here the nature of the uncertainty will be manipulated by the perceptual features on the screen and the amount of experience DMs are allowed to have with the particular or related objects. Experiments within these paradigms will investigate, as a function of the nature of the uncertainty, how attention is allocated to the different components of a decision problem, how DMs combine their prior beliefs with new information, and how they combine judgments of differential precision and reliability. In addition to the theoretical role of this research in understanding human decision behavior and factors that affect its optimality, it is relevant to situations in which human DMs are required to make choices at least in part on the basis of indirect information and considered judgments by other people. Space exploration, nuclear power plant sitings, and the disposal of hazardous wastes are just a few examples of such situations. Findings from this research may allow the development of more optimal communication and decision procedures. In addition, the research will assist in the development of computerized decision aiding expert systems that attempt to represent human understanding of linguistic variables and vague information.
