From the lab to the cockpit: the development and analysis of quantifiable measures of goal-directed actions through gaze behaviour

Abstract

Objectives: The goal of this thesis was to examine the utility of gaze behaviour as an indicator of information processing during the performance of complex tasks in ecological environments. Information processing is influenced by the interaction between task requirements, an individual’s resource availability, and the environmental context (i.e., modifying factors). Five experiments were performed that manipulated these modifying factors and examined the respective changes across low fidelity and high-fidelity environments. Study one comprehensively examined gaze behaviour while participants performed a neurocognitive test where task demands increased progressively by loading the working memory system (i.e., cognitive load). Studies two and three characterized gaze behaviours associated with increased sensorimotor task demands during the performance of simulated landing scenarios using a computer-based simulation and high-fidelity flight simulator, respectively. Study four aimed to explicitly assess differences in gaze behaviour between novices (i.e, ab-initio) and experts (i.e., licensed low-time pilots) during a simulated flight task, while study five expanded on findings from these studies by examining gaze behaviour during an extended (i.e., cruise and landing) flying scenario that involved a dual-task paradigm (i.e., cognitive load).

Methods: Twenty-seven participants were recruited for study one. Task performance and eye movements were recorded as they completed the Tower of London task, which progressively increased in cognitive load by manipulating problem set size (i.e., minimum number of moves to solve the task). In studies two and three, sensorimotor difficulty was manipulated by altering the wind conditions during a landing scenario (i.e., no wind, turbulent winds). Eighteen participants in study two and twenty-nine participants in study three were recruited and had their gaze behaviour recorded during the task in a low-fidelity (i.e., PC-based) and high-fidelity (i.e., fully immersive flight training device) simulation environment, respectively. Study four examined gaze behaviour and performance differences during a simulated landing task between ab-initio pilots (i.e., 0 flight hours) and licensed low-time pilots (i.e., on average 193 flight hours). Study five examined secondary auditory oddball performance, gaze behaviour and flight performance as a function of task condition (single, dual) and flight phase (cruise, landing).

Results: Results from the first four studies revealed several uses for traditional (i.e., dwell time, fixation rate, fixation duration, scan path length) and advanced (i.e., SGE, GTE, cognitive tunneling analysis) gaze behaviour metrics. Specifically, when task requirements increased (i.e., sensorimotor demands, cognitive load) the focus of attention became more biased toward areas that held more task-critical information and were necessary for successful task performance. Many of these measures were also sensitive to phase specific changes in gaze behaviour associated with distinct task goals. Indeed, overall shifts in gaze dispersion, dwell time, and scanning patterns were shown to reflect sources of information that were important for achieving flight task specific sub-goals. Such differences were evident independent of changes in sensorimotor difficulty or cognitive load. These changes in gaze behaviour due to variation in task requirements were consistent across low-fidelity and high-fidelity cockpit environments. Several gaze behaviour measures also demonstrated that experienced pilots used vision more efficiently and effectively, by sampling more task-related regions of the environment more frequently with fewer instances of cognitive tunneling. The last aviation study was the only paradigm that did not show significant differences in gaze behaviour or flight performance due to increased cognitive load. This was despite the presence of increased auditory response times and error rates. Although most gaze metrics examined here were modulated (or influenced) by a number of factors impacting skill performance, blink rate seemed to be the one measure that was not consistent.

Conclusions: Gaze behaviour analysis provides important spatiotemporal information related to the underlying cognitive processes supporting complex skill performance. Gaze behaviour metrics are sensitive indicators of various task requirements and individual resource availability. It is important to note that the gaze behaviours reported are highly task dependent and are stable across the environmental contexts examined here. Collectively, the gaze metrics examined here provide valuable information for assessing information processing and can contribute to better characterization of visual scanning in complex environments.