Expressing and interpreting the results of quantitative risk analyses. Review and discussion

One of the most fundamental problems when using quantitative risk analyses is related to the way the results of the analyses are expressed and interpreted. Should we restrict our attention to unobservable quantities such as hazard rates and statistically expected values (frequencies), or should we focus on the prediction of observable quantities such as the occurrence of an accidental event, the time to a specific event occurs, etc.? In this article we review the alternative approaches and discuss the implications concerning e.g. the meaning of ‘uncertainty’ of the risk numbers generated. We conclude that in practice there is no alternative but adopting the latter approach, focusing on observable quantities and using ‘subjective probabilities’. The subjectivistic (Bayesian) theory of probability provides the framework for the coherent use of judgement, which constitutes a significant (sometimes the only) part of the information that is available to us. However, it is not clear to most risk analysts what the Bayesian approach really means. There is a strong need for a more attractive presentation of the Bayesian approach. This article intends to contribute to such a presentation.