An intelligent decision support system for decision-making based on vague, imprecise, and incomplete information

Prescriptive decision theory requires decision-makers to express their beliefs and tastes with explicit, precise, and complete information. This information, expressed in subjective probability and utility functions, is often difficult and costly to obtain from a decision-maker. It is obvious that this information requirement is much too demanding in most practical applications, and often unreliable and biased, thus inappropriate for use in traditional decision analysis. However, an optimal strategy in a decision problem is generally robust to the precision of the subjective probability and utility information. In RID, the only information elicited is that which the DM wishes to provide, to whatever degree he or she is willing and able to do so. Only strong binary preferences are expressed for alternatives or partial/full decision functions to minimize inconsistencies. These judgments and preferences are incorporated into an interactive procedure which repeatedly prunes the probability and decision space until a single "optimal" or a small subset of efficient decision functions remains. Risk analysis with an imprecise utility function is also investigated. Agents with different preferences may agree on comparisons and even problem solutions at the same time, and individuals with preferences which are intermediate with respect to two individuals who agree on a comparison may not agree with the other twos. A multicriteria methodology is proposed to enhance the versatility of RID which takes advantage of the structure of AHP and solid theoretical aspects of MAUT, in addition to the robustness of RID. It enables a decision-maker to build an evaluation hierarchy to decompose a complicated problem into a number of independent modules. We enhance the capability of RID/MCRID with embedded intelligence to make them distinguishable from other DSS's. Two main issues are concerned. The first, based on the behavior of the decision-maker and his/her understanding on the problem itself, performs strategic prediction when the input information is not sufficient to determine the optimal solution. The second deals with knowledge and/or behavior accumulation and its reuse; based on this behavior consequences can be used to estimate possible optimal solution or set of feasible alternatives. In either case, they speed up information processing under uncertainties. (Abstract shortened with permission of author.)