Errors in technological systems

Massive data and experience exist on the rates and causes of errors and accidents in modern industrial and technological society. We have examined the available human record, and have shown the existence of learning curves, and that there is an attainable and discernible minimum or asymptotic lower bound for error rates. The major common contributor is human error, including in the operation, design, manufacturing, procedures, training, maintenance, management, and safety methodologies adopted for technological systems. To analyze error and accident rates in many diverse industries and activities, we used a combined empirical and theoretical approach. We examine the national and international reported error, incident and fatal accident rates for multiple modern technologies, including shipping losses, industrial injuries, automobile fatalities, aircraft events and fatal crashes, chemical industry accidents, train derailments and accidents, medical errors, nuclear events, and mining accidents. We selected national and worldwide data sets for time spans of up to similar to200 years, covering many millions of errors in diverse technologies. We developed and adopted a new approach using the accumulated experience; thus, we show that all the data follow universal learning curves. The vast amounts of data collected and analyzed exhibit trends consistent with the existence of a minimum error rate, and follow failure rate theory. There are potential and key practical impacts for the management of technological systems, the regulatory practices for complex technological processes, the assignment of liability and blame, the assessment of risk, and for the reporting and prediction of errors and accident rates. The results are of fundamental importance to society as we adopt, manage, and use modern technology. (C) 2003 Wiley Periodicals, Inc.