This preview shows page 1. Sign up to view the full content.
Unformatted text preview: y functions to form a multiattribute utility function. This is also done using
simple lotteries to reveal indifference between receiving a particular set of attribute values with certainty, or, a
lottery of attribute values. In its simplest form, the resultant multiattribute utility function is a weighted sum of the
individual attribute utility functions.
Evaluate each alternative using the multiattribute utility function. The most difficult problem with MAUT is getting the decision makers or evaluators to think in terms of lotteries. This can
often be overcome by an experienced interviewer. MAUT is based on a set of mathematical axioms about the way
individuals should behave when confronted by uncertainty. Logical consistency in ranking alternatives is assured so long
as evaluators adhere to the axioms; no guarantee can be made that this will always be the case. An extended discussion
of MAUT is given in Keeney and Raiffa, Decisions with Multiple Objectives: Preferences and Value Tradeoffs, 1976. A
textbook application of MAUT to a NASA problem can be found in Jeffrey H. Smith, et al., An Application of Multiaffribute
Decision Analysis to the Space Station Freedom Program, Case Study: Automation and Robotics Technol ogy Evaluation,
1990. studies in which the alternatives cannot be described by
a set of continuous mathematical relationships.
Selection Rules When Uncertainty Predominates.
When the measures of system effectiveness,
performance or technical attributes, and system cost for
the alternatives in the trade study look like those for
alternative C in Figure 22, the selection of the best
alternative may need to be handled differently. This is
because of the general propensity of decision makers to
show risk-averse behavior when dealing with large
variations in cost and/or effectiveness outcomes. In
such cases, the expected value (i.e., the mean) of some
stochastic outcome variable is not a satisfactory point
measure of that variable.
To handle this class of decision problem, the
system engineer may wish to...
View Full Document