Modelling pilot decision-making errors in New Zealand general aviation : a thesis presented in partial fulfilment of the requirements for the degree of Master of Arts in Social Sciences at Massey University

Abstract

Accident statistics indicate that the rate of mortality and financial loss associated with general aviation accidents is comparable with that of passenger transport operations. However, general aviation appears under­ represented in literature pertaining to the development of safety interventions.

In this thesis, this apparent disparity is addressed in an investigation of pilot error in New Zealand general aviation. Using the precedent of accident modelling developed in industrial safety research, accident models taken from aviation, road transport and industrial settings are reviewed for their representation of human error. The Surry Model (1969), a twelve point sequence representing operator decision making processes, was selected for generalization to aviation.

The selection of this model was congruous with research literature identifying poor decision making as a primary causal factor in air accidents. Each of the points in the model represents an opportunity for accident avoidance if certain information processing requirements are met.

The model presents accident avoidance as the result of three processes: the correct recognition of stimuli, the correct cognitive processing of avoidance options, and the correct implementation of physiological responses. The accident sequence within which these processes occur is divided into two cycles: the build-up of danger in the system, and its subsequent release.

The model was applied to a data base of 84 cases involving fixed wing aircraft engaged in general aviation, selected from 1980 to 1991. The point at which an error in pilot decision making occurred was identified and coded using the twelve points of the Surry Model. These data were combined with information concerning biographic characteristics of the pilots, and the number of passengers on board the flight. All pilots in the sample were male. Two research questions were investigated. The first questions whether the Surry Model is a useful tool in the analysis of information about accident sequences. The model was used as a template, and laid over the time line of accidents, as they had been determined by air accident investigators. The second research questions sought to determine whether the format of the model could be used as a protocol for developing time lines and questioning pilots during accident investigations.

A small final sample size resulted in a general dichotomizing of the variables for non-parametric Chi Square statistical analysis. The power and utility of the analysis was limited and could only show that, beyond chance effects, there were no biographic characteristics of pilots that influenced the cycle of the model in which the accident inducing error occurred.

No quantitative examination of the twelve error types identified by the model was possible. A low level of inter-rater reliability showed that the model was not as self-contained as anticipated. Raters appeared to use the model in a consistent manner, but modes of use varied between individuals. It is suggested that this may be a function of non-standardised presentation of human factors information in air accident reports, coupled with non­ standardised interpretations of ambiguities in the model.

On the basis of the inferential interpretation of the data, two main areas of discussion arise. The first is concerned with 'ambiguities': the structural characteristics of the Surry Model that influenced the fall of data onto the twelve error types. It became apparent that the typical sequence of events in aircrashes differed from the temporal sequence depicted by the model, and that assumptions made in the model about the configuration of the pilot- aircraft interface were inaccurate. Accordingly, modifications to the model are proposed.

The second area of discussion is centred on 'antidotes': corrections for pilot errors identified as causal in aircrashes. The results indicate that some aspects of in-flight behaviour could be targeted for intervention. It is suggested that it may be useful to encourage pilots to engage in active information search from external sources in order to ensure that they supplement information available from the aviation system. Self-monitoring before flight may induce voluntary self removal from aviation activities. It is possible that some pilots may abstain from flight if they become aware that their performance has become impaired as a result of their physical or emotional condition.

It is also suggested that risk communication techniques could facilitate the development of worst case thinking by pilots who are confronted by potential hazards. Rather than a more traditional emphasis on the implementation of strategies after contact with danger, these antidotes may encourage the active avoidance of danger.