Based on detailed accident data gathered on the scene by multidisciplinary teams (440 situations studied) and carried out using system approach logic, this study presents a framework of human error analysis, applied to the field of car driving. This framework was based on traditional error analysis models derived from cognitive and ergonomic psychology (e.g. Rasmussen, Reason), adapting them to the specific features inherent in the activity of driving and its corresponding difficulties. These features include: complexity, variability of situations, weak formal definition of procedures, time and dynamic constraints, and the crucial role of anticipating what is about to happen. It accounts for some twenty functional errors involved at different stages of the performance of the activity (perception, diagnosis, prognostication, decision-making, execution of the action). The framework also relies upon the description of accident mechanisms in the form of standard scenarios. These scenarios are aimed at explaining the regularities found in the performance process and the contexts which generate these errors, including task dysfunction, endogenous and exogenous explanatory factors, and the resulting critical situation. Using ergonomic reasoning, we also present a set of application studies that use an analysis of accidental errors in order to diagnose driver aid needs and the conditions for meeting them. This deals with the qualitative definition of needs extracted from clinical data, their quantitative assessment based on a representative sampling of police accident reports, an estimation of the capacity of aid devices (Prometheus) to meet these needs, and a definition of the potential limits to the use of these aids. (Author/publisher)
Abstract