Irrational Economist_ Making Decisions in a Dangerous World - Erwann Michel-Kerjan [125]
A related situation occurred in the early 1990s, during a discussion of the potential risks associated with the Ghost Dance Fault that crosses through the designated nuclear repository site at Yucca Mountain in Nevada. I was working with a select group of seven seismologists who were the world’s experts on that fault; they were also familiar with probability assessment. A key issue was whether the fault was active. I said to them, “Without communication with each other, write down your judgment of the probability that the fault is active.” Then, before any discussion, I asked each to write down the definition of active, which is one of the most basic words in earthquake science. Everyone was surprised when we got six different definitions. Some of these referred to what happened on the fault in the past, others concerned what might possibly happen in the future, and still others concerned what would definitely happen in the future.
One of the seismologists then suggested that it would be interesting if they each wrote down their definition of the Ghost Dance Fault. Upon reviewing the definitions, we were astounded to find that we had two Ghost Dance Faults. The problem stemmed from the fact that a north-south fault segment ran from north of the repository southward through the repository to where it was covered up by a large landslide several thousand years old. There was also a smaller fault segment that ran under that landslide from the southwest. It turned out that some of the seismologists thought both segments were part of the same fault, which they referred to as the Ghost Dance Fault. Others felt that the Ghost Dance Fault was only the larger north-south segment. Clarifying that there were different definitions of the Ghost Dance Fault greatly facilitated our communication about the possible likelihood and magnitude of any earthquake on the fault.
The point of all these examples is that characterizing a policy decision using only general terminology is ineffective and can even be misleading. Yet the exclusive use of such general terms is often how policy issues are discussed. The decision sciences fields have long known that carefully defining terms and quantifying likelihoods clarify the meanings of qualitative terms and reduce ambiguity in many situations. This knowledge would be useful in the policy arena as well.
ASSESSING, UNDERSTANDING, AND COMMUNICATING ABOUT UNCERTAINTY
The problem of interpreting and manipulating probabilistic information is complex and likely the source of important misinterpretations on major policy problems. The following two examples give support for this assertion.
In 1988, I was a member of a working group sponsored by the U.S. Geological Service to specify probabilities of major earthquakes occurring on segments of various California faults. The other eleven members of the committee had substantial experience with and knowledge about California earthquakes. My role was to help them come up with internally consistent sets of probabilistic estimates. A major earthquake was defined as a Richter magnitude 6.5 or greater. Five segments were defined on the two major faults in the San Francisco area, namely the San Andreas Fault and the Hayward Fault. A 45-kilometer length of the San Andreas Fault was called the Loma Prieta segment. In the course of several meetings, the probability that a major earthquake would occur over the next thirty years was assessed as 0.3 for the Loma Prieta segment and as 0.2 separately for each of the other four segments. On October 17, 1989, a magnitude 7.1 earthquake on the Loma Prieta segment occurred, leading to significant damage