Information and confidence levels in risk results: Can both be obtained?

Validation and confidence in risk analysis results are difficult to achieve for Low Probability–High Consequence (LPHC) events. By definition, such events do not provide empirical evidence and hence expert judgments are used in deriving probability measures regarding the occurrences and consequences of such events. The more complicated the risk analysis is in terms of model structure, parameters, variables, and assumptions, the more judgments are needed. The user of the risk analysis results may rightfully ask ‘what is the confidence in the results?’ In the paper a linkage between a variable’s information- and confidence-level is shown where a higher information-level is coupled with a lower confidence-level for LPHC events. This suggests one way of dealing with confidence: a risk analysis result may be presented with lower information but with higher confidence. If such a trade-off is not possible, or does not solve the problem regarding confidence, basic management strategies related to ‘low’, ‘medium’ and ‘high’ confidence levels are proposed to guide the decision-maker. The paper ends with the conjecture that a three-level confidence scale is enough for grading the confidence in risk analysis results related to LPHC events.