Models for assessment of human error in system reliability studies.

Abstract

The probabilistic analysis of risks applied to complex industrial systems demonstrates that human error is a very important part of the total risk, although its exact quantification is doubtful. On the other hand, the current techniques for modeling and calculating human errors presuppose, in a way not always explicit, the election of a certain pattern of human behavior; in such a way that many limitations revealed by the scientific literature seems due to the use of so named ‘partial’ behavior models, while the ‘general’ models, that are more ambitious, still need better methodological development. The present work describes the methods applied currently to human reliability studies and it points out the relationship that underlies between these and the models for human behavior representation, showing the investigation lines ongoing for covering methodological lacks. We declare the protagonism of Psychology about defining the human behavior models that underlie the reliability analysis techniques, and also about criticizing these techniques, especially on the simplifications that, to make them more governable, they have the perverse effect of hiding psychological mechanisms of errors.