One of the traditional approaches to reducing risk is to work on reducing the severity (S) of an accident rather than reducing the Likelihood (L) of occurrence. As the classical definition of risk is, Risk (R) = L X S, by reducing S we reduce the risk. Simple really…
However, reducing severity also serves another purpose in that it addresses the completeness problem of any risk analysis. For example how do we really know that our analysis has identified all the relevant causal factors for a hazard. If we haven’t caught them all then we will have under estimated the likelihood of the hazard occurring and therefore resulting in an accident, which then becomes an often overlooked (ontological) component of risk.
As by definition we don’t know what these causal factors are the logical conclusion is in any system where there is a high degree of ontological risk we should emphasise severity reduction rather than trust the completeness of our safety analyses over-much.
Authors Note – 1 Aug 2012: In this and other posts I’ve been somewhat loosely using the term ‘epistemic’ to denote both epistemic and ontological uncertainty. So I’ve revised this post to more clearly differentiate between ontological and epistemic uncertainty.