ABSTRACT
When the Berkeley ‘high-reliability organisation’ (HRO) group first set out to examine the role of operators in maintaining safety in a variety of complex, high-technology operations, the intention was to produce a reasonably ordinary, organisational study, essentially positivist in orientation and based on a combination of behavioural analysis, survey research instruments, and empirical observation (La Porte, 1996; Rochlin, 1993). My HRO colleagues and I sought to link what operators believe, or say they believe, about their performance and the performance of their systems with observations about operational ‘reliability’ that involve measurable indicators. In some cases, such as aircraft launches from an aircraft carrier, indicators of this sort could be derived if the data were made available. In others, such as nuclear power plant operations or air traffic control, it soon became clear that there was no baseline for measuring failures. Moreover, and more surprisingly, it became clear that, even if such data were available, they would not provide indicators of operational safety. Paradoxically, operators’ perceptions of potential risk were an essential element in their construction of an environment of safe operations (Roberts, 1993; Rochlin, 1996). Operational safety is more than a culture. Even within the organisation itself safety is a social construct. It is therefore difficult not only to measure but to analyse and characterise with the usual analytic tools of organisational theory. It involves élan and modesty, pride in past accomplishments and concern about future ones, cooperation between operators and technical cadres and also tensions between them, as well as an acknowledgment of the importance of close oversight by managers and regulators, and a demand for independence of action. The sense of accomplishment is defined as a blend of autonomy, safety, and production goals (Roberts, Rousseau, and La Porte, 1993). But what sort of blend? And to what extent can a constructed belief in operational safety be measured against some set of observations or empirical measures, say, of reliability or of the control of risk?
