Modeling operator performance for error analysis

Existing human performance taxonomies which mostly can answer the question "what factors could affect the result" rather than "why it happened", are usually used to analyze the accident reports, and fail when they are applied to investigate the underlying causes. This is because the existing taxonomies provide modeling primarily for latent conditions and not so much for active errors which can often trigger an accident. In the paper, we present the method for operator error modeling, which describes active error at the "sharp end" of the system, and together with the existing classification taxonomy for latent conditions constitutes a powerful modeling tool for human performance analysis. This method, which was implemented in aviation, is demonstrated for the analysis of the pilot\´s errors. It can be used as a basis for developing human performance algorithms in complex man-machine systems to improve operator training, assessment, error analysis, and investigation.