Accident causation models, management and the law

To apportion blame, and by extension liability, for an accident it is necessary to decide causality, who caused the accident and how it was caused. The same requirements apply to the preventative management of such potential accidents, except blame is assigned post-hoc, after the event, whereas preventative management is essentially proactive and obviates the need for blame. Much thinking is based on the notion that there is a single root cause of an incident, the most important cause and therefore the one pointing at liability as well as determining the main target for prevention. This is embedded in the idea that incident causation is linear and deterministic, that there are clear sequences of causes going back to a root cause. This way of thinking has proved very successful and its preventative application may be regarded as reducing the number of (potential) accidents by 80%. Most of these 80% accidents are personal; the development and use of the Swiss Cheese model, aimed also at process incidents, has led to a further reduction of possibly 80% of the remaining potential incidents, now covering some 96% in total. Such models are still deterministic, but non-linear in their causal effects. The remaining 4% of possible incidents, especially complex and major process accidents, unfortunately appears to be much more intractable. The proposal is that these incidents have a causal structure that is both non-linear and non-deterministic, being inherently probabilistic. This has consequences for the management and prevention of such incidents, because of their complexity, but also for the legal approach, that has to confront non-deterministic and non-linear causation. The legal viewpoint is made more complex because, in hindsight, such incidents still appear to be simple, linear and deterministic.