The concept of cognitive workload can be defined in very general terms as the amount of mental resource a person needs to utilise to perform a particular task (or range of tasks) in a given environment. The definition implies that people have a limited amount of cognitive resource available and that mismatches between environmental demands and the availability of cognitive resources is a common cause of human error in work systems.
Whilst there is strong evidence to suggest that workload levels are indeed related to the occurrence of human error, research has also shown that the link between external demand and available internal resource is not straightforward. For example, whilst high levels of mental workload tend to give rise to “cognitive strain” a condition in which the human information processing system is unable to cope with the large amounts of environmental stimuli competing for limited attentional resources, a reciprocal effect obtains when environmental demand is low. In this situation, people tend to become less vigilant and attention begins to wander increasing the probability of missing important information available for selection in the stimulus array. The tendency for quality of task performance to be hindered by strain (e.g., excessive demand) or boredom (e.g., insufficient demand) can be characterised with reference to Yerkes-Dodson law (Yerkes and Dodson, 1908). This law proposes that optimal task performance occurs at an intermediate level of mental arousal (i.e., workload), with much poorer performance levels resulting from lower and higher arousal levels. When plotted in a graph, Yerkes-Dodson leads to a predicted inverted U relationship between arousal and performance quality. Unfortunately, the precise shape of the inverted U function has been found to vary with changes to the nature of the task. Thus, optimal performance tends to occur at much lower levels of arousal than would be predicted by the ideal Yerkes-Dodson curve when tasks are easier. Conversely, higher levels of arousal are needed for effective performance of more difficult tasks.
A further complication relating to the development of a better understanding of cognitive workload arises relative to the finding that arousal/performance
can vary widely within a given population and can even be different for the same person performing identical tasks on two separate occasions.
2.1 Workload and error
The complexity of the relationship between workload, task performance and task load can be illustrated with reference to the debate in which a number of investigators have aimed to provide an answer to the question “How much workload is too much?” (e.g., de Waard, 1996; Meijman and O’Hanlon, 1984; Teigen, 1994). To answer this question, investigators have found it useful to divide the Yerkes-Dodson inverted U function into 6 task performance-related regions as shown in Figure 8.
Figure 8 Task performance and workload as a function of demand
(Adapted from de Waard, 1996)
2.1.1 Optimal Performance (A2)
When task demand is in the A2 region, the human operator is able to cope easily with workload and performance remains at levels approaching optimal. Moderately increased demands do not lead to significant increase in cognitive strain and extrinsic factors do not unduly affect performance. If errors do occur they tend to arise due to factors other than those associated with task demand or cognitive workload.
2.1.2 Increased Workload Demand (A3-B-C)
When task demands increase to levels within the A3 region, measures do not typically show any noticeable decline in performance despite increased task
loading. The operator is only able to maintain adequate task performance levels, however, by increasing cognitive effort – or depending upon the theory preferences of the analyst, by allocating more mental resources to processing activity. Limited amounts of time spent working in the A3 region are not thought to be detrimental to the well being of the human operator and fall within normal operating parameters. Extended periods of time spent in the A3 region, as might occur say when peak loads are frequently experienced in the work environment, can lead to emergence of potentially harmful stress effects which are to be avoided wherever possible (e.g., Mulders et al, 1988). The likelihood of a person experiencing stress is greater when the operator has no control over the work conditions giving rise to heightened workload.
With further increases in demand, performance levels transition to region B. Here the quality of task performance begins to decline because workload demands start to exceed the operators’ capacity to cope. Again, dependant upon the type of model of workload being used, this can either mean that a limited-capacity channel has become fully occupied with a particular processing activity, or alternatively, it could mean that available cognitive resources are insufficient to meet the current task demand profile. Whichever is the case, in this region performance errors become increasingly commonplace as task demand increases. Individuals no longer have the mental resources needed to recover the situation without adopting coping strategies – which in some situations could involve reducing demand by jettisoning some of the work activities contributing to cognitive overload.
Finally, when workload levels exceed the threshold for entry into region C the operator is at risk of losing control of the situation due to high workload levels and may begin to experience extreme effects of psychological stress. In region C, performance levels are at their lowest levels and can only be improved by reducing workload demands.
2.1.3 Decreased Task Demands (A1-D)
In contrast to the high workload scenario, reduced task demands from the A2 region transition into region A1 in which maintenance of performance relies heavily on increased operator vigilance. It should be noted that, contrary to popular belief, vigilance is a cognitive activity, which implies increased – rather than decreased - cognitive workload levels for reasons that will be outlined in due course. Acceptable levels of vigilance can only be maintained in the A2 region by increased cognitive effort, although in this case, the effort needs to be directed towards the maintenance of a vigilant state, keeping the cognitive system in a state-of-readiness for response, rather than directing attention toward incoming task demands as in the case of high workload situations (Caggiano & Parasurraman, 2004; Warm, et al, 1996).
occurs, errors of omission become prevalent in human performance, as the human information processing system is unable to maintain arousal levels sufficient to ensure the registration of the target events whose identification are necessary to trigger the appropriate response. Interestingly, the transition from region A2 to D tends to occur below the level of conscious awareness, which means that the operator is largely unaware that they are no longer attending to the vigilance task.
2.2 Workload assessment and system safety
The division of the inverted U function into six ‘task performance’ regions provides a valuable qualitative classification useful for discussing cognitive workload in practical contexts. In particular, the classification provides a means of establishing when safe upper and lower limits of workload demand have been transgressed, an activity which has proved especially problematic for most workload assessment tools. Using the classification described above, error producing workload conditions arise when workload demands exceed the regions A2-B threshold. Conversely, the risk of vigilance errors becomes greatly increased when task demand falls below threshold values which define the A2-A1 boundary. Unfortunately, the classification is unable to provide insight into how the analyst might define the performance boundaries in absolute terms using task performance measures from real world situations.
2.3 Background Summary
Despite all the problems associated with the concept of mental workload, it is widely recognised that there are significant benefits to be gained from the ability to make estimates of mental workload levels in particular situations. Possible uses for a workload assessment tool in hazardous environments are:
• To identify unsafe error producing conditions in normal or abnormal work situations
• To evaluate deployment of new equipment (i.e., automation) • To assess manning levels in high-hazard environments • To evaluate impact of proposed changes to work methods • As an aid to incident investigation