On surveys of expert opinion

Probabilistic risk assessments are, like most computational assessments, subject to knowledge uncertainties. Quite often experts are consulted for their opinion on the subject matter in question to bridge the corresponding gaps in the computational assessment procedure. The less to the point information available, however, the more the expert judgement depends on features that may vary considerably from expert to expert. Additionally, a number of biases are pointed out in the literature that may have serious impact on expert judgements. Some of these biases are outlined and the variation in opinions between experts is illustrated by practical examples from different engineering disciplines. With respect to the credibility and usability of PRAs to nuclear power plants it is concluded that if expert opinion is to serve as a significant quantification basis it is to enter in form of a survey of expert opinion. (orig.)