Multiple Objective Evaluation and Choicemaking under Risk with partial Preference Information.
Abstract
In this paper, we consider a single stage, multiattribute decision making problem under risk. Each outcome is associated with a vector of value scores. The set of value score vectors is partially ordered in the usual manner. This partial order is used to induce partial orders on the set of alternatives based on: expected value score, first order stochastic dominance, and second order stochastic dominance. We show that properly aggregating value scores produces a more refined partial order on the set of value score vectors, which then induces a more refined partial order on the alternative set. This result forms the basis of a decision aiding procedure which: (1) identifies the nondominated set of alternatives, a set guaranteed to contain the most preferred alternative under mild assumptions, (2) asks the decision maker to choose a most preferred alternative from the nondominated set, and (3) if this choice cannot be made, aggregates some, but not necessarily all, of the value scores to strengthen the partial order on the alternative set, thus reducing the size of the nondominated set and presumably enhancing alternative selection. The potential value of this procedure is that it is interactive, it can accommodate a variety of levels of online preference feedback from the decision maker, and it does not necessarily require that the value scores by completely aggregated. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1981
- Accession Number
- ADA097418
Entities
People
- Andrew P. Sage
- Chelsea C. White Iii
Organizations
- University of Virginia