Thinking and Deciding

MAUT AS A TYPE OF DECISION ANALYSIS 355
you would respond to pregnancy. Would you have an abortion if you or your partner
were pregnant now? Or would you plan to have the baby?
The next step is to compare each of the other attributes to the one that is most
important in terms of its range. A simple method is to ask yourself, “How big, in
terms of its effect on what I care about, is the difference between the top and bottom
of the smaller range, compared to the difference between the top and bottom of the
larger range?” This gives you the weights. In Table 14.2, for example, a couple
(presumably) has judged HIV prevention as what matters most, with sexual pleasure
equal to that. (They would be indifferent between abstaining and taking the risk of
HIV infection.) And they judged the range for other STDs as mattering half as much.
To carry out a consistency check on the weights, a simple method is to use some
other attribute than the most important one as the standard. For example, use pregnancy
prevention as the standard. Then easy use should have a weight of .50 rather
than .40, since pregnancy prevention itself has a weight of .80 relative to HIV prevention:
.50 · .80 = .40. This kind of consistency check is analogous to a check for
ratio inconsistency (p. 339). If the consistency check fails by a large amount, try to
figure out why and make adjustments accordingly.
Complete the analysis by multiplying the attribute utility by the weight and summing
across rows. For condom, the sum is 1.00 · 99 + .50 · 99 + .80 · 84 + .40 · 0+
1.00 · 90 = 305.7.
One interesting feature of this analysis is the treatment of HIV risk. Rather than
defining the bottom end of the dimension as getting AIDS for sure, this analysis
uses an estimate of the risk of HIV infection for a particular person. If the analysis
had defined the bottom of this dimension as AIDS for sure, then the weight of this
dimension would have to increase factor of 1,000. But the values of the various
methods on the dimension would then range from 99.9 to 100 instead of from 0 to
100.
Test your understanding. Would it matter if the values ranged from 0 to .1 instead
of from 99.9 to 100 (assuming that abstinence was always assigned the highest
value)? 6 What would we do with someone who had the same values but a much
lower risk of AIDS? 7
Decision analysis is often applied to decisions that are difficult because the options
appear to be close in utility or expected utility. This appearance is often correct.
A disturbing fact of life is that options are often equally good, insofar as we
can determine their goodness, even when the decisions are important and when great
uncertainties are present. Von Winterfeldt and Edwards (1986) refer to such situations
as “flat maxima” because, when expected utility is plotted as a function of
some attribute (for example, the amount of money invested in a project), the function
typically does not have a sharp peak: Within a fairly large range, the value on the
attribute does not make much difference. Once we determine that we are in such a
hypothetical and are identical in all other respects.
6 No. It would just lower all the utilities by a constant.
7 If the risk were .0001 instead of .001, then we could either divide the weight by 10 or have the values
range from 90 to 100 instead of 0 to 100.