When we look at the evolution of accident investigation over the last century, we see a development from seeing accidents as:
1. failures by individuals who need to be retrained or replaced;
2. active failures influenced by latent conditions within work systems, which need to be better engineered, manufactured and controlled;
3. influenced by the unpredictable nature of work environments with their networks of influences, which have to be supported in order to anticipate and adapt to unexpected situations.

The quote from Karl Popper illustrates what we often do in safety management: capture reality by imposing categories and theories on it. For instance, our Safety Management System requires we select one "basic cause" of "basic risk factor".

Karl Popper taught us to make predictions that can be tested, and reject our theory if these predictions are shown not to be correct. Yet, safety practitioners are really busy fulfilling company duties. Regulators look at the requirements rather than at the evidence that something works. How can we work together and collect data on safety practices, in order to know better what works in practice?