This preview shows page 1. Sign up to view the full content.
Unformatted text preview: more detailed study.
AHP assumes the existence of an underlying
preference “Vector” (with magnitudes and directions)
that is revealed through the pair-wise comparisons.
This is a powerful assumption, which may at best hold
only for the participating evaluators. The figure of
merit produced for each alternative is the result of the
group’s subjective judgments and is not necessarily a
reproducible result. For information on AHP, see
Thomas L. Saaty, The Analytic Hierarchy Process,
1980. NASA Systems Engineering Handbook
Systems Analysis and Modeling Issues
Multi-Attribute Utility Theory
MAUT is a decision technique in which a figure of merit (or utility) is determined for each of several alternatives
through a series of preference-revealing comparisons of simple lotteries. An abbreviated MAUT decision mechanism
can be described in six steps:
(5) (6) Choose a set of descriptive, but quantifiable, attributes designed to characterize each alternative.
For each alternative under consideration, generate values for each attribute in the set; these may be point
estimates, or probability distributions, if the uncertainty in attribute values warrants explicit treatment.
Develop an attribute utility function for each attribute in the set. Attribute utility functions range from 0 to 1; the
least desirable value, xi0, of an attribute (over its range of plausible values) is assigned a utility value of 0, and
the most desirable, xi*, is assigned a utility value of 1. That is, ui(xi0) = 0 and ui(xi*) = 1. The utility value of an
attribute value, xi, intermediate between the least desirable and most desirable is assessed by finding the value
xi such that the decision maker is indifferent between receiving xi for sure, or, a lottery that yields xi with
probability p i or xi* with probability 1 - pi. From the mathematics of MAUT, ui(xi) = pi ui(xi ) + (1 - pi) ui(xi*) = 1 - pi.
Repeat the process of indifference revealing until there are enough discrete points to approximate a continuous
attribute utility function.
Combine the individual attribute utilit...
View Full Document