excerpt

edit

Excerpt from Clemen 2001:

"Much of the difficulties in decision when different people have different ideas regarding some aspect of the decision. The solution is to refine the conceptualisations of events and variables associated with the decision enough so that it can be made. How do we know when we have refined enough? The clarity test (Howard 1988) provides a simple and understandable answer. Imagine a clairvoyant who has access to all future information: newspapers, instrument readings, technical reports, and so on. Would the clairvoyant be able to determine unequivocally what the outcome would be for any event in the influence diagram? No interpretation or judgement should be required of the clairvoyant. Another approach is to imagine that in the future, perfect information will be available regarding all aspects of the decision. Would it be possible to tell exactly what happened at every node, again with no interpretation or judgement? The decision model passes the clarity test when these questions are answered affirmatively. At this point, the problem should be specified clearly enough so that the various people involved in the decision are thinking about the decision elements in exactly the same way. There should be no misunderstandings regarding the definitions of the basic decision elements.

The clarity test is aptly named. It requires absolutely clear definitions of the events and variables. In the case of the EPA considering toxic substances, saying that the exposure rate can be either high or low fails the clarity test; what does “high” mean in this case? On the other hand, suppose exposure is defined as high if the average skin contact per person-day of use exceeds an average of 10 milligrams of material per second over 10 consecutive minutes. This definition passes the clarity test. An accurate test could indicate precisely whether the level of exposure exceeded the threshold.

Although Howard originally defined the clarity test in terms of only chance nodes, it can be applied to all elements of the decision model Once the problem is structured and the decision tree or influence diagram built, consider each node. Is the definition of each chance event clear enough so that an outside observer would know exactly what happened? Are the decision alternatives clear enough so that someone else would know exactly what each one entails? Are consequences clearly defined and measurable? All of the action with regard to the clarity test takes place within the tables in an influence diagram, along the individual branches of a decision tree, or in the tree’s consequence matrix. These are the places where the critical decision details are specified. Only after every element of the decision model passes the clarity test is it appropriate to consider solving the influence diagram or decision tree."

References

edit

Howard, R. A. (1988). Decision Analysis: Practice and Promise, Management Science, 34, 678-695

Clemen, R. T. & Reilly, T. (2001). Making Hard Decisions with DecisionTools Suite, p. 77