Deriving human-error tolerance requirements from tasks

Shows how an understanding of a dynamic system from the point of view of the tasks that it supports and an understanding of human error can guide a process of deriving human error tolerance requirements. Our aim is to provide a means whereby, rather than relying on training as a means of improving operator performance, designers may develop interactive systems with human error tolerance in mind. We extend an established methodology, the systematic human action reliability procedure (SHARP), by employing a software engineering notation, communicating sequential processes (CSP), that provides a bridge between error theory and the practice of design and implementation. In this paper, we outline approaches to human error, describe a task notation based on CSP which helps us to elicit requirements on human-error tolerance expressed as functional properties of the system. The technique is used to analyse an engine fire recovery procedure in order to derive human error tolerance requirements