Human Factors in the Design of Tactical Display Systems for the Individual Soldier (1995)

There are many arguments in the research literature on how to define cognitive workload. We begin, for pragmatic purposes, with a simple definition:

time required/time available

If the time required by a task is more than the time available, we define it as cognitive overload. If the time required is very much less than the time available (and there are no other tasks that a soldier has to do), we define it as cognitive underload. The major problem with the time-based approach is the question of parallel processing. That is, individuals--especially, skilled individuals--can do more than two things at once. As a result, tasks overlap, and it is difficult to provide a concrete estimate of time required since tasks performed at the same time can appear to take “no time.”

Some tasks, such as extended surveillance, require a soldier to pay attention for long periods of time without overtly doing anything. Although this represents an underload situation, it can be stressful and, almost paradoxically, can result in cognitive overload. Therefore, the cognitive workload scale is best seen as a continuum in which too much and too little are liable to result in problems.

Solutions to the cognitive overload include training: a skilled soldier can deal with more mental demands than his unskilled counterpart. However, the preferred option is to structure information in order to adapt to the capabilities of less skilled soldiers. This approach has the advantage of dealing with the underload problem since certain information can be presented when little else is going on. As an active soldier's existence is, like that of several other professions, described as “hours of boredom and moments of terror,” the adaptive approach is likely to prove more efficient and cost-effective then extensive training, although training is still a necessary component of operations.

There are several mental workload measurement methods. Their respective advantages and disadvantages have been the subject of extensive scrutiny (see Gopher and Donchin, 1986; Hancock and Meshkati, 1988; Hancock et al., 1985; Hart and Staveland 1988; Meshkati et al., 1989; Moray, 1979; Wilson and O'Donnell, 1988). For the sake of simplicity, we have summarized the major pros and cons of each method in Table B-1. (See Lysaght et al., 1989, for an evaluation of method with respect to Army programs.)

The traditional methods by which workload has been assessed include engineering formulations, such as time-line analysis, and relatively simplistic representations, such as the time required by a task as a function of the time available. These quantitative approaches

assume certain characteristics about human operators, principal among which is stationarity. If humans worked in the same way as certain engineering or electrical systems do, such approaches would be acceptable and effective in many operations. However, human operators are not made in this fashion.

Indeed, one central characteristic of human operators is their nonstationarity: that is, they can and often do process in parallel (as noted above), and they grow fatigued over time, but, conversely, they learn readily in some situations. Furthermore, people are able to take nonlinear leaps of intuition that cannot be captured by simple linear causative models of systems operation. In consequence, the hopeful but naive mathematical expressions of workload, while they might represent a first-pass approximation, largely fail to capture the subtleties and nuances of experience of people as they grapple with the varying demands of complex tasks.

In contrast to mathematical formulations, there are psychophysiological approaches to the assessment of cognitive workload. Between these two forms of measurement methodologies lies the principal form of assessment, primary task performance.

If it were not for the fact that one wishes to predict workload, simply measuring the output efficiency of individuals with respect to the task they are performing would provide, in large part, the required information. But because one wants to predict when and how individuals might encounter conditions that exceed their capability that more than primary task measures are needed. However, this does not mean that primary task performance cannot be used to reflect more than instantaneous load level. For example, if a soldier had performed a sequence of operations before, primary task measures might indicate periods of stressful operations. As mentioned above, there is always the question of context. In aviation for example, it is still reasonable to suggest that periods of high workload are most likely around take-off, landing, or emergency conditions. In the same fashion, if a soldier's mission can be evaluated in terms of expected load, primary task measures can still be most useful, especially if training or simulation of the event is possible beforehand. For actual, on-line measurement of experienced load, primary task performance is a critical source of information to both central battle managers and potentially, to soldiers themselves.

Attention and cognitive workload are intimately linked psychological constructs. With the postulation of Kahneman's concept of attentional resources, there came a paradigm shift in the area of attention that changed the perspective on attention from a filter or gate function

to an energetic representation. Obviously, the latter construct is far more amenable to integration with a workload perspective. Intrinsic to the original resource formulation was the postulate that attention was a unitary resource of limited capacity. By inference, tasks that were performed concurrently (dual-task performance) competed for such limited resources. The demands of performing one task were consequently reflected in the performance level of a second, or secondary task. For those who tied workload directly to attention, workload on a main task could be measured as a reflection of performance level on this subsidiary or secondary task.

Today there are a number of objections to this form of assessment procedure. With respect to soldiers' tasks, the first, pragmatic, and most important objection is that the introduction of a secondary, often artificial, task distracts a soldier from primary performance. Given the presence of other measures of workload, this is a needless addition. In any case, the argument for attention as a unitary resource structure began to fall apart under empirical attack; although hybrid offspring were proposed, they “explained” findings only at the expense of ever greater “explanatory” degrees of freedom. With respect to the assessment of cognitive workload of a soldier, the secondary assessment method of workload measurement is not particularly useful.

While objective measures of what a soldier is actually doing are helpful to battle managers (and perhaps to the soldier), it is important to integrate the affective domain of response. Since the proposed technologies are to be used under imminent life-threatening situations, a critical component of the context is the subjective reaction of a soldier. (As often noted, even weapons are not used by a significant percentage of soldiers in the heat of battle.) If technology is to prove effective, some assessment of a soldier's subjective state is important.

There are several well-known methods of subjective workload assessment. Typically, they elicit paper-and-pencil responses for subsequent analysis, although this is not necessary. Like measures of primary performance, subjective responses represent, principally, the momentary reaction of the individual, although some work has been done on retrospective subjective assessment. The advantage of subjective ratings is their high face validity, which is a nontrivial factor when dealing with professional operators. However, current assessment procedures are cumbersome, and there is still disagreement about the necessity for initial baseline procedures (such as the weightings factor in the NASA Task Load Index (TLX) (Hart and Staveland, 1988) and the card sort procedure in the Subjective Workload Assessment Techniques (SWAT). Indeed, it may well be the case that individuals do provide retrospective assessments of the difficulty of particular situations, but in a common professional jargon. It might be possible to link formal subjective techniques to such informal observations to provide useful on-line information concerning the subjective state of individual operators. Indeed, informally this process would already be a part of the

battle manager's task in assessing the state of both the troops and ordinance under their control.

One of the great hopes in mental workload assessment is the development of a nonintrusive technique that taps directly into cognitive function. The ability to provide information on the direct status of cognitive abilities presents a formidable challenge, but one with extensive payoffs if successful. Like other psychophysiological techniques, the general form of inquiry is a search for a relationship between some measurable facet of central nervous system activity and some concomitant behavioral phenomenon. The more global the behavior specified (e.g., sleeping versus waking) and the more widespread the physiological spectrum of inquiry, the more liable that linkage can be made.

Since cognitive workload is a function of effort associated with a task, there is reason to believe that psychophysiological techniques should at least provide some hope for assessment. Individuals have searched almost the whole spectrum of reasonable psychophysiological functions--and even several that are not--to find a measure. The most promising general candidates are related to heart function; however, like all such techniques, it is a matter of distilling signal from noise when one is not sure what a signal looks like. Often those signals are weak, need amplification, and are subject to unfortunate artifacts. For example, evoked potentials have always held particular promise, but they are accompanied by a number of problems (see Humphrey and Kramer, 1994). In laboratory settings, psychophysiological equipment can be somewhat intimidating in terms of size and configurations, but contemporary progress in biomedical engineering makes the possibility of micro-assessment systems a firm probability. That these could be easily incorporated into head-mounted displays is a systems question for future design. Should such integration be possible, psychophysiological measures represent a strong candidate for inclusion in a battery of cognitive workload assessment techniques.

If one retains a perspective in which attention is linked, at least at some level, to attention, one can begin to construct a picture of what is liable to happen to workload as extensive training is practiced. In related empirical work, Hancock (1984) has argued that skilled individuals experience less workload at a common task demand than their untrained peers. This view is founded on a link to automated processing; it is also used to support a similar contention about training and stress effects.

Briefly, training on consistent elements of high-performance skills permits the development of automated responses. In general, these responses are appropriate since the relation between input (task) stimuli and output (response) actions does not change over time. For example, the military strategy of overlearning skills such as unjamming a rifle would fall into this category. Automated responses, unlike “control” or effortful responses, require a minimal investment of attentional resources. They can be performed with great speed and can be done in parallel with other operations. Although these pristine effects are probably constrained to the research laboratory, something very similar does occur in the actual operational environment. As tasks become fully automated, the workload associated with them drops. Thus, training becomes a very important facet of operations.

What remains problematic is the context of operations. An important experimental question that has begun to be addressed is how context affects overlearned performance skills. The implication from laboratory research is that context is not absolutely critical. However, for infantry soldiers, it may well be that context of operation is the overriding factor. Resolution of these two differing views will dictate whether training on task or training on task in context is preferred. In supporting the latter, it critical to note that how a task is structured and displayed may be the dominant influence on the contextual effect. Given that head-mounted displays can present virtual as well as real environments, the “displayed” context may be controllable; however, few soldiers will suspend reality totally if they are being shot at. The question of context then assumes great importance in the training and workload relationship.

Yet another issue is that of individual differences. There has been relatively little information on individual differences in the mental workload domain. Damos (1988) in her review of the area, could only find six studies that dealt with factors associated with individual differences and experienced workload. Unsurprisingly, there were no indications of general trends, and few conclusions could have been drawn since such studies mostly dealt with different aspects, such as personality traits or behavioral patterns. It might be possible to extend the impoverished understanding of individual differences by looking at the literature on other “energetic” constructs, such as stress, fatigue, anxiety, and attention.

There are many unresolved problems concerning the assessment and interpretation of mental workload and its demands. One is the unknown effects of combining physical and mental demands was mentioned above. However, there are many other questions that have to be faced in using workload assessment. As just noted, for example, there is very little

information on differences between individuals (Damos, 1988). The effects of training on mental workload are still to be clarified, especially when training takes place in conditions that only simulate actual operational environments (see Hancock et al., 1994). The effects of task failure on workload are uncertain (Hancock, 1989), and strategies to improve performance through adaptive or compensatory systems have only now begun to provide potential answers (Chignell and Hancock, 1985; Hancock and Chignell, 1987; Hancock et al., 1994). Also unknown are the relationships between stress, workload, and other energetic aspects of performance, such as situational awareness. In particular, there is concern that under certain conditions, workload and performance dissociate so that workload increases but performance gets better or, similarly, workload decreases but performance worsens (Derrick, 1988; Yeh and Wickens, 1988). This is especially disturbing for those who want to use cognitive assessment of workload as a basis for design decisions or the definition of operational procedures. As technical support systems are developed, it becomes vital to address these experimental questions.