Technological Accidents: Learning from Disaster

Financial support from the program " Risques collectifs et situations de crise " of CNRS is gratefully acknowledged. In 1979 a partial melt-down of the reactor core at Three Mile Island was the worst nuclear power accident to that time. massive radiation overdoses during treatment using the Therac-25 machine. In the summer of 1991, ten million subscribers in the United States lost telephone service for durations ranging from ten minutes to six hours. In 1996, after ten years of development and a $500 million investment, the first Ariane 5 rocket exploded shortly after takeoff. In each of these cases, technological failure occurred. Loss of life, loss of huge investments, or loss of consumer services, caused disasters in a broad sense. In each case, the source of the disaster was uncovered and steps were taken to ensure that it did not re-occur. The disaster was followed by learning. The purpose of this paper is to investigate that learning, and to understand the economic aspects of " learning from disaster ". From the point of view of the economics of knowledge, a disaster, like many other phenomena, is an opportunity to produce knowledge, to reduce the probability of future disasters, and to improve the performance of technological systems in general. 1 The opportunity lies in the fact that a disaster can be seen simply as an unplanned experiment producing data about the system concerned. In principle, these experiments were always possible — the European Space agency could deliberately have exploded a rocket — but the costs (broadly defined) of performing such an experiment were extremely high. Ex ante, expected costs exceeded expected benefits, and the experiment was not run. But ex post a disaster has the effect of lowering the cost of the experiment to zero — the experiment has been performed. The issue is then to profit from the experiment, and thus simply to maximize cognitive, technological, and organizational benefits. Few empirical studies explicitly address the economics of technological disasters as learning opportunities. Studies of disasters tend to focus on what, or how much was learned, and they do show that learning tends to follow a disaster. For example, the gas spill in Bhopal and Three Mile Island have both been examined to see the consequent changes in operation, organization and reliability. The Bhopal chemical disaster changed the way the chemical 1 For recent attempts at qualifying learning as the production …