Citation

Baker DP, Krokos KJ. Hum. Factors 2007; 49(2): 185-199.

Affiliation

American Institutes for Research, 1000 Thomas Jefferson St., NW, Washington, DC 20007, USA. dbaker@air.org

Copyright

(Copyright © 2007, Human Factors and Ergonomics Society, Publisher SAGE Publishing)

DOI

unavailable

PMID

17447662

Abstract

OBJECTIVE: This investigation sought to develop a reliable and valid classification system for identifying and classifying the underlying causes of pilot errors reported under the Aviation Safety Action Program (ASAP). BACKGROUND: ASAP is a voluntary safety program that air carriers may establish to study pilot and crew performance on the line. In ASAP programs, similar to the Aviation Safety Reporting System, pilots self-report incidents by filing a short text description of the event. The identification of contributors to errors is critical if organizations are to improve human performance, yet it is difficult for analysts to extract this information from text narratives. A taxonomy was needed that could be used by pilots to classify the causes of errors. METHOD: After completing a thorough literature review, pilot interviews and a card-sorting task were conducted in Studies 1 and 2 to develop the initial structure of the Aviation Causal Contributors for Event Reporting Systems (ACCERS) taxonomy. The reliability and utility of ACCERS was then tested in studies 3a and 3b by having pilots independently classify the primary and secondary causes of ASAP reports. RESULTS: The results provided initial evidence for the internal and external validity of ACCERS. Pilots were found to demonstrate adequate levels of agreement with respect to their category classifications. CONCLUSIONS: ACCERS appears to be a useful system for studying human error captured under pilot ASAP reports. Future work should focus on how ACCERS is organized and whether it can be used or modified to classify human error in ASAP programs for other aviation-related job categories such as dispatchers. APPLICATION: Potential applications of this research include systems in which individuals self-report errors and that attempt to extract and classify the causes of those events.

Language: en