Escaping the Non-Quantitative Trap

Quantitative modelling and analysis is common in safety engineering, but it is often criticised. Objections include the difficulty in acquiring probabilities (e.g. for human error), the dubious assumptions often needed to manipulate them (e.g. independence of events), and the inherent uncertainty involved in making decisions based on probabilistic predictions. Clearly, poor predictions are of little value and may be dangerous. Faced with this danger, many people respond by eliminating quantities altogether. This is a trap, as we have no guarantee that the resulting model or predictions will be better; indeed, the subtlety of expression offered by numerical probabilities has been lost. This paper discusses some alternatives to the non-quantitative trap, and explores their significance for the issue of safety case assurance.