Dealing with uncertainty is a difficult problem; decision analysis relies upon
probability theory to capture the uncertainty faced by the decision maker. In
the engineering of a system the uncertainty is not often described by existing
data and interpretable as the long-run frequency of a set of known events.
Instead the uncertainty deals with processes that change with time and
for which no (or at most a few) known events have occurred. Instead of
ignoring the uncertainty faced in the engineering of a system, decision analysis
permits the engineers to capture the expert judgment of the engineers,
stakeholders, and other experts and use this information to provide insights
about the design choices with the best information available at the time. Recent
advances in decision analysis provide graph-theoretic models for representing
probabilistic dependence (relevance diagrams) and decisions problems (influ-
ence diagrams).
Once uncertainty is modeled explicitly, the risk preference of the decision
maker has to be addressed as part of decision analysis. The concepts of risk
aversion, neutrality, and preference are defined mathematically and illustrated
as part of the decision analysis process. Using the decision maker’s risk
preference requires computing the certainty equivalent as the inverse of the
utility function.
Clearly, it is inappropriate to use the sophisticated tools of decision analysis
for every decision that is part of the engineering of a system. Many times
engineers have described the benefit of thinking about the decision in the terms
of decision analysis. At other times developing the value model and using a
quick scoring and weighting evaluation provides insight into which alternatives
are serious and which should be ignored. For really complex and contentious
decisions, the full power of decision analysis can provide an explicit and
rational process for defining and discus sing the alternatives to reach a
conclusion consistent with the values of the stakeholders and the uncertainty
as defined by relevant expert s.
PROBLEMS
13.1 In defining reliability of a system, we talk about the probability of a
failure. Failure here is an event or distinction, but not one that passes
the clarity test. As a result, systems engineers work very hard to focus
on the distinction, mission failure, where a mission failure is a failure
that precludes the user from completing her/his mission. This definition
still does not pass the clarity test because we have not defined the
mission, a definition that is system and context dependent.
For the elevator system wher e you work or go to school,
a. Define mission in a way that meets the clarity test.
b. Define as many failures as possible and show which would be
classified a mission failure. Be sure to keep the clarity test in mind
when defining these failures.
PROBLEMS 443