In both an applied and theoretical setting, the understanding of the human factors that contribute to differing levels of cognitive workload is imperative to managing most safety critical work (Djorkic, Lorenz &Fricke, 2010; Neal et al., 2013). There are various cognitive mechanisms underlying workload, which Neal et al. (2013) broadly define as the psychological state an individual enters into when interacting with a task, requiring demands of the task relative to the performers capacity. Research is especially important in this area, since the cost of errors on the job are potentially high (D’Arcy & Rocco, 2001; Wilson, 2001) and findings suggesting that not only work overload, but work underload can influence overall performance; safety performance; and employee health (Muth, Moss, Rosopa, Salley & Walker, 2012; Neal et al., 2013). In terms of performance, Lysaght and colleagues (1989) indicate that the relationship between performance and workload form an inverted ‘u’, such that low levels of workload can lead to boredom: which is related to missed signals; instructions; and/or decreased performance. At high levels of workload, it has been found that performance is also negatively influenced, due to the high volume of competing demands (Lysaght et al., 1989). Researchers indicate that an optimum level of workload leads to acceptable performance (Lysaght et al., 1989). Overall, the research field is aimed at mitigating the risk of mental strain and its adverse effects, to ultimately provide comfort in working conditions for employees (Nickel & Nachreiner, 2003).
Over the last five decades, various theories and models of workload have been proposed, however none of these effectively measure or predict workload (Loft, Sanderson, Neal, Mooij & Murphy, 2007; Neal et al., 2013). Air traffic in Australia is growing, such that for all Australian airports the number of passengers has gone from 9,031,447 in February 2009 to 12,959,196 in July 2014 (Australian Government, Airport traffic data). Evidently air traffic is on the increase, existing air traffic management systems will be put to the test (D’Arcy & Rocco, 2001) and thus the most sophisticated systems are required in order to manage air traffic and those who control it. In order to achieve this industry goal, the objective of the present research is to gain a better understanding of Air Traffic Controller (ATC) workload. This research is aimed at deciphering what workload looks like operationally, including the human factors that contribute to variability in workload. This will be achieved by coupling heart rate variability (HRV) and self-report data in a holistic manner, as opposed to focusing on the negative impacts of overload/underload or potential error. Additionally, we can also start including analysis on how a human operates and add value to the overall system performance.
The research aims and objectives align with that of the safety II philosophy, which has recently gained research attention in industry (Hollnagel, 2014). The safety II philosophy is different to that of the classic safety I, such that safety I denotes that errors occur because of technical; organisational; or human causes. Investigation and safety audits only occur when an incident or error occurs. However, a shift has been seen in the study of safety, where safety II focuses not just put on what has gone wrong, but what has gone right in the system. This shift in focus sees the whole system being monitored, not only those sections that have been problematic and led to incidents (Hollnagel, 2014; Hollnagel, Leonhardt, Licu & Shorrock, 2013).
Air Traffic Controller Workload
The present research was conducted on en route ATC sectors: this is a portion of the airspace approximately 48 km from an airport (Loft et al., 2007). The roles and responsibilities of an en route ATC involves accepting aircrafts; checking aircrafts; issuing instructions; anticipation of loss of separation between aircrafts; intervening to resolve conflicts; advice to pilots; and handing over an aircraft to another sector where applicable (Loft et al., 2007). Literature has put forward various factors as contributors to the complexity and difficulty of an ATC role, such as aircraft count; sector geometry; traffic flow; separation standards; aircraft performance characteristics; and weather (Kopardkar & Magyarits, 2003). Neal et al. (2013) indicate that these factors have been found to account for approximately half of the variance in workload. However, there is still a large proportion of variance that is not accounted for (Neal et al., 2013).
Looking to workload modelling, various attempts have been made to predict workload levels to manage high workload periods (e.g. by splitting sectors) (Loft et al., 2007). Theories put forward by Sperandio (1971) suggests that the association between task demand and cognitive workload is a function of the ATC management strategies. Management strategies then form a feedback loop back to the ATC. Loft et al., (2007) built on this theory of workload prediction indicating that the ATC management of resources is important to the relationship between task demand and workload.
Assessment of Workload
Within the aviation industry and human factors, various measurement methods have been utilised in order to report and quantify an individual’s perceptual and physiological response to cognitive workload. These methods have been tested on a variety of roles, including ATC and pilots during a number of field studies and experimentally manipulated simulator studies (Matthews et al., 2014; Muth et al., 2012; Neal et al., 2013; Nickel & Nachreiner, 2003).
In terms of self-report measures, the NASA-TLX is a popular questionnaire developed by Hart and Staveland (1988). The measure consists of six subscales (mental demand, physical demand, temporal demand, effort, performance and frustration) to obtain an overall measure of workload (Muth et al., 2012). Muth et al., (2012) outline the limitations of subjective measures like the NASA-TLX, which is based on the conscious response that the operator needs to produce a rating of their perceptions of workload at a given time. Although the NASA-TLX measure is popular, there are some researchers who have questioned the methodology of this measure since its development (de Waard & Lewis-Evans, 2014). More specifically, other measures have been found to be more sensitive (Rating Scale Mental Effort) and the scales and items had been merely adopted from previous workload measures without an empirical basis.
For the present study, a self-report measure was developed from the scale in Air Traffic Workload Input Technique (ATWIT) utilised by Stein (1985). This scale was designed on the basis that the operator experiences workload and they are able to provide estimates (on an anchored scale) of their performance relative to the level of work they are completing for both high and low workload levels (Stein, 1985). This modified scale also obtains a rating of each factor that has been suggested to be important in influencing workload, and the relative importance of each given factor in driving levels of workload.
It is generally accepted that self-report measures cannot measure workload alone (De Waard & Lewis-Evans, 2014), as research indicates they should be used in conjunction with physiological indicators of workload to obtain a holistic overview of workload (Muth et al., 2012). In terms of physiological workload measurement, a variety of methods have been utilised over the years. These range from Electrocardiagram (measuring of electrical activity when the heart contracts), Electroculography (measuring of eye movements), Electroencephalography (electrical activity in the scalp), blink rate and Heart Rate Variability (Wilson, 2001). The study of HRV is a popular method in the workload literature, as it has been consistently linked to correlates of cognitive workload (Lean & Shan, 2012; Matthews et al., 2014). Heart Rate Variability is also thought to be the most suitable measure of workload (Strang, Best & Funke, 2014). According to Muth et al. (2012), HRV is controlled by the autonomic nervous system (ANS), which has been shown to be associated with workload; the sympathetic nervous system (SNS) and parasympathetic nervous system (PNS) also influence HRV. Respiratory Sinus Arrhythmia (RSA) is a fluctuation in HRV related to respiration across a number of frequencies and has been established to differentiate between changes in levels of workload (Muth et al., 2012). In terms of RSA output, the higher the level of workload, the lower the RSA frequency as the PNS activity is subsiding (Muth et al., 2012).
Respiratory Sinus Arrhythmia (RSA) has been consistently shown to be highly associated with cognitive workload, as it is objective and empirically validated (Muth et al., 2012; Nickel & Nachreiner, 2003). More specifically, it was found that for Navy aviation personnel, measurements of RSA was significantly associated with self-report measures of workload (using the NASA-TLX) and is suggested to be capable of differentiating between tasks that are deemed high or low workload (Muth et al., 2012). A study conducted by Splawn and Miller (2013) using volunteer civilian and military personnel found that HRV was able to differentiate between different task loads. It was also found that performance and heart rate measures can be utilised in order to match that of perceptions of cognitive workload. Added to which, the study conducted by Strang et al. (2014) found a significant relationship between cognitive workload and heart rate metrics, such that heart rate was a significant predictor of cognitive workload above and beyond any other predictors entered. However, interesting results were found for different task types. This indicates that there are unique differences in the tasks presented to participants (Strang et al., 2014) and thus possible underlying task demands that HRV may be more sensitive to.
Research conducted by Nickel and Nachreiner (2003) suggests that HRV was able to differentiate between work and rest, however was not sensitive enough to detect changes in levels of workload under controlled laboratory conditions when compared to subjective measures of workload. It is also suggested that RSA may be influenced by emotional strain (Hockey, Nickel, Roberts & Roberts, 2009; Nickel & Nachreiner, 2003). Strang et al. (2014) suggest that there is a high degree of complexity and sometimes inconsistency in the results pertaining to the measurement of HRV on workload, indicating that further research is required in order to understand the existing relationship and ultimately inform workload modelling.