The wrong Wright stuff : mapping human error in aviation

The Aviation Safety Reporting System (ASRS) was instituted to aid the Federal Aviation Administration in tracking trends in aviation incidents so that, ultimately, safety measures and training could be implemented to decrease the occurrence of accidents and incidents within the industry. The current system relies on hand coding of reports to recognize current trends and alert the proper parties. Although the filing party may enter some codified data describing the surrounding scenario (e.g., time of day, weather), there is no opportunity to specify a category if the problem is human error. Considering the prevalence of human error within these incidents (around 55% based on a report by Boeing, 2006), a greater understanding of the driving factors is needed. The current study was an investigation of the human error components of airline incident reports. Text analysis tools were applied to ASRS incident narrative reports to determine a classification based on human performance for commercial and general aviation. The results from the current study demonstrate that an empirically based approach can be used to uncover latent categories within the Flight Crew Human Performance\u27 classified reports. The combined approach of latent semantic analysis, k-means clustering, and keyword analysis were used successfully in developing a nine element classification of commercial aviation reports and twelve element classification of general aviation reports. The taxonomies suggested by the current study for both commercial and general aviation reveal categories beyond just human error elements. The classification scheme suggested for the commercial aviation reports most closely resembled the ACCERS taxonomy developed by Krokos and Baker (2005; see also Baker & Krokos, 2007), which was constructed to help in categorizing all incident reports. The classification suggested for general aviation reports did not closely resemble any existing classification scheme. Although the suggested taxonomy shared categories such as situational awareness and communication with classifications such as crew resource management (CRM) or single pilot resource management (SRM), the current classification also holds non-human elements such as weather and context. The taxonomies for both commercial and general aviation revealed a category for context, and the difficulty of flying into certain airports was apparent. These findings can be implemented to improve training programs by assisting in the creation of contextually based training scenarios. Furthermore, based on findings for general aviation in particular, pilots could benefit from increased training in situational awareness and monitoring of notices and airspace