02.03.2013 Views

Thinking and Deciding

Thinking and Deciding

Thinking and Deciding

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

346 DECISION ANALYSIS AND VALUES<br />

of psychology might make it harder to get a majority vote for a c<strong>and</strong>idate directed<br />

more at “wisdom <strong>and</strong> enlightenment” (2C). When I realized this, I also realized that<br />

we had other options for dealing with this problem, such as changing the voting rules<br />

so that the scientific <strong>and</strong> wisdom parts of the department would have equal say, no<br />

matter how many members each area had. We did not do this, but my realization<br />

illustrates one of the beneficial side effects of this kind of analysis: it leads us to<br />

consider a wider range of possible options, <strong>and</strong> sometimes one of these will be better<br />

than the options in the range we were considering.<br />

Decision analysts sometimes convince a group of people to use a framework<br />

like this to make a real decision. The process is time consuming, too much for<br />

the purpose at h<strong>and</strong>. It would require several steps. The initial diagram would be<br />

developed by the analyst after preliminary discussion, then presented to the group<br />

for their criticism. Everyone should agree that it includes all relevant goals, however<br />

important or unimportant they seem to each person. The idea is that people will agree<br />

on this list even if they disagree on the weight that each element should receive. This<br />

process can lead to increased agreement. It can force each member to give some<br />

consideration to goals that are important to others but that she would ignore. She<br />

would ignore them because of her single-minded concentration on only the goals<br />

most important to her. A listing of attributes is this an antidote for single-mindedness.<br />

The next step would be to define each attribute so that it can be measured numerically.<br />

Then the group might be able to agree on the assignment of numbers to each<br />

option (in this case, each field in which the department hire, perhaps) on each attribute.<br />

This can be done by averaging individual estimates. Such agreement is more<br />

likely when people agree on the meaning of the attributes. No agreement about their<br />

importance or weight is needed yet. Finally, each member would assign a weight to<br />

each attribute. An analysis for that member would multiply the weight by the number<br />

assigned to each option on each attribute <strong>and</strong> add up across the attributes so as<br />

to assign a utility to each option. (Later, I shall give more detailed examples of this<br />

process of multiplying <strong>and</strong> adding.) At this point, the analyses of individual members<br />

would be aggregated into an overall analysis. This could be done by averaging<br />

the weights assigned to attributes or (equivalently) the utilities assigned to options.<br />

Ordinarily, this would be done in a preliminary way, then discussed by the group,<br />

then done again.<br />

Conjoint measurement<br />

Before returning to multiattribute analysis, I want to step back to examine its theoretical<br />

basis in conjoint measurement. Conjoint measurement is a mathematical idea<br />

that justifies multiattribute analysis <strong>and</strong> that has also led to another method for utility<br />

assessment, called “conjoint analysis,” which is discussed in Chapter 15.<br />

Conjoint measurement deals with the situation we have been discussing, in which<br />

outcomes can be described in terms of two or more attributes. An “attribute,” in this<br />

case, is a value on a dimension. The price of a given computer, for example, is an

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!