Investigating the under-reporting of automation dependency in air transport accidents worldwide

Introduction. The International Civil Aviation Organization (ICAO) tasked the Personnel Training and Licensing Panel (PTLP) Automation Working Group with conducting a study concerning management of automation, manual flight, and pilot monitoring. The tasking was a high priority given the international interest in vulnerabilities to increasing application of automated systems in air transport flight decks. Part of the Study was to determine the scope of automation dependency in air transport aircraft worldwide. Automation dependency is when the pilots accept what the aircraft automation is doing without adequately monitoring or confirming that the aircraft is doing what they expected or wanted it to do. Automation dependency is also commonly referred to as automation over-reliance or automation complacency within existing literature. One of the key findings of the Automation Study Report, documented in ICAO Circular 361, was that automation dependency goes under-reported in accidents and incidents. Research question. This analysis explored whether automation dependency was identified when indicators were present in air transport aircraft accident reports. Method. Twenty-eight air transport accident reports, that were determined to demonstrate automation dependency by the ICAO Automation Working Group, were systematically reviewed for confirmatory identification of automation dependency by the investigation report. Results. 6/28 (21%) of the reports explicitly stated automation dependency or other common terminology within the report. 1/28 (4%) of the reports included language that alluded to automation dependency. The majority 21/28 (75%) of air transport accident reports reviewed did not identify the presence of automation dependency when indicators were present. Discussion. Accurate tracking of instances of automation dependency in accident and incident reports is challenging due to inconsistent identification of the phenomena within reports. As the accident reports that were reviewed demonstrated indicators of automation dependency, it was expected that the reports would contain language that identified it as such, though the majority did not meet this expectation. There may be benefits to improving appropriate identification of automation dependency reporting when indicators are present to ensure cases are adequately addressed, tracked, and potential mitigations developed. Conclusion. Given that the majority of air transport accident reports reviewed did not identify the presence of automation dependency when indicators were present, it is likely that automation dependency is under-reported in operational flying, which impairs the veracity of tracking the construct within the field.