On the nature of risk and accidents

By consensus, 'errors of omission' and 'errors of commission' have their origin in the human rather than in the technological parts of the system. It has therefore been customary to base explanations on the available psychological 'error models', which mostly have been of the human information processing and internal 'error mechanisms' type. Yet the fact that an 'error of commission' necessarily involves a human action does not mean that the explanation, hence the modelling, must be confined to human functions - cognitive or otherwise. On the contrary, the experience from accident analyses as well as state-of-the-art HRA approaches all emphasize that action failures reflect the impact of the working conditions more than any specific mechanism'. This corresponds to a change in accident models from sequential, over epidemiological, to systemic. According to the latter, action failures - or action error modes - should be seen as the irregular, but not totally random, outcome of the system's performance variability rather than as specific effects of specific causes. This change in perspective have significant consequences for the ways in which action error modes are modelled and classified, and therefore also for the methods that can be used for prediction. (author)