Abstract

Situation awareness (SA) has been the subject of a heated debate since its inception, and periodic examination of important concepts that undeniably have significant impact on human factors work is valuable. Endsley (2015) has certainly provided a thorough contrast between her 1995 three-level SA model (Endsley, 1995) and several more recent models to better distinguish the variety of views on the concept of SA. Parasurman, Sheridan, and Wickens (2008) have also methodically documented how SA is distinct from other constructs, such as attention, memory, and mental workload; from other processes, such as decision making; and from other performance, subjective, or physiological measures. A point that we take as the starting point of our present comments is Parasuraman et al.’s (2008) recommendation that theories in cognitive engineering should be held to the standards of being able to specify the constraints in which they account for more or less variance in human–system performance and thereby “successfully [predict] performance and [make] testable recommendations for the improved design of systems” (p. 155).
In this paper, we argue that the use of the SA construct in conjunction with that of mental workload can further help specify the constraints and account for variance in human–system performance, especially that of heavily automated systems. More specifically, we argue that the extent to which the SA and mental workload measures associate or the manner by which they dissociate can provide valuable insights concerning the human–system operational efficacy. Later in the paper, an emerging approach that attempts to use both workload and SA measures to engage adaptive system changes will be described.
The potential relationship and distinction between SA and mental workload emerged early in the development of the SA concept (e.g., Endsley, 1993; Wickens, 2001). In previous work (Vidulich, 2003; Vidulich & Tsang, 2012; Wickens & Tsang, in press), we have provided a conceptual sketch on the dynamic interaction of the two constructs. Generally, mental workload has been most often characterized in terms of the level of attentional demands placed on the operator in the course of performing required tasks, whereas SA is primarily associated with the informational content of the operator’s memory systems during task performance. Both the level of workload and the quality of SA are shaped by exogenous and endogenous factors. Exogenous factors include operational demands, situation complexities, and system constraints. Endogenous factors are inherent in the operator’s ability and skill. For example, the same operational demands and a similar level of situation complexity would impose higher workload and likely bring about poorer SA for a less skilled operator than for an expert operator. There are both theoretical arguments and empirical support (e.g., Just, Carpenter, & Miyake, 2003) for the notion that high-ability or high-skill operators effectively have more resources at their disposal. It is also clear that experts are adept at using their vast amount of domain-specific knowledge in long-term memory strategically to circumvent the severe capacity and temporal limits of their working memory (Ericsson & Kintsch, 1995).
Inasmuch as mental workload is caused by and SA is supported by many of the same cognitive processes, both are subject to the limits of and compete for many of the same processes. The more demanding and complex the task, the more work is required to perform the task and to assess the situation. When a greater amount of attention is required to support performance, less is available for monitoring, interpreting, comprehending, and updating information in keeping abreast of the situation. In this case, a high level of mental workload could lead to poor SA. On the other hand, SA could be improved by working harder by, for example, more frequent information sampling. That is, a high level of workload is sometimes necessary to maintain good SA.
Because mental workload and SA could support each other as well as compete with each other, knowing about only one of them is often not sufficient for assessing the effectiveness of task allocations, team structure, interface design, or such. One assessment technique that was developed to assess both workload and SA is the situation present assessment method (SPAM; Durso & Dattel, 2004). An additional advantage of this technique is that its SA queries and latency-based workload measures can potentially be obtained in real time, providing additional insights into the operators’ strategies for managing performance, workload, and SA (Durso & Alexander, 2010).
Another potential real-time technique is the Global Implicit Measure (GIM), which was developed with the expressed aim of obtaining real-time SA measures to guide automated pilot aiding (Vidulich & McMillan, 2000). After a detailed task analysis, mission-specific behaviors are identified and scored during mission operations. Moment-by-moment comparison of achieved versus expected or optimal goals is made. Failures to accomplish mission goals are indicative of the operator’s being either unaware or unable to perform at the moment. Initial validation of the technique showed that it was able to distinguish successfully between two different cockpit designs and different mission phases.
Whereas both the SPAM and GIM techniques rely on performance measures, an emerging neuroergonomic approach makes use of neurophysiological and performance data to inform about the operator’s mental workload and SA in real time to guide the implementation of aiding in advanced human–automation systems in real time. For example, Galster and Johnson (2013) and Parasuraman and Galster (2013) proposed a “sense, assess, and augment” framework to guide system adaptation to the human operator’s needs. The functional state of the operator will first need to be sensed via behavioral or neural measures for identifying information overload. These measures will then be assessed relative to certain performance, mission, or system requirements. In conjunction, neurophysiological techniques (such as functional magnetic resonance imaging and/or event-related potentials) would be used to identify the level of activity in key brain areas. Noninvasive brain stimulation (e.g., transcranial direct current stimulation) could potentially be applied to the relevant cortical regions to improve performance.
The development and implementation of closed-loop adaptive operator aiding, such as that represented by the sense-assess-augment framework, is likely to be greatly aided by invocation of the operator’s SA and mental workload constructs. For any given task environment, it is likely that some combination of sensors and assessment algorithms will be appropriate for detecting whether the operator is experiencing excessive mental workload (i.e., the overtaxing of available attention resources for processing task demands) or poor SA quality (i.e., inappropriate information-seeking behavior or maladaptive neural activation). Whatever is sensed and assessed in terms of SA and/or mental workload difficulties could then inform decisions about the appropriate aiding. For example, the optimal aiding for an operator who is attending to the appropriate information sources and showing the expected neural activations (suggesting good SA) but also showing physiological reactions indicative of information overload (suggesting mental workload challenges) would probably be very different from that for an operator showing the reverse pattern, even if the level of performance were the same for both operators.
In the spirit of Parasuraman et al.’s (2008) suggestion, the development of real-time adaptive human–machine systems are predicted to be another excellent test bed for the utility of the SA and mental workload concepts to guide human factors practice. All of the preceding research in these fields provides ample justification for being optimistic about their utility. But time, and research, will tell.
Footnotes
Michael A. Vidulich is a research psychologist in the Applied Neuroscience Branch of the Air Force Research Laboratory’s Human Effectiveness Directorate. He previously served as the technical advisor for the Warfighter Interface Division and the Collaborative Interfaces Branch. With a PhD in experimental/engineering psychology and more than 30 years’ laboratory experience, he specializes in cognitive metrics for human–machine interface evaluation and adaptation. Some of his significant accomplishments include creating real-time capable performance-based situation awareness metric the Global Implicit Measure and developing the Subjective Workload Dominance metric. He also demonstrated the use of workload redline procedures in Air Force Test and Evaluation and helped develop the NASA–Task Load Index rating scale. He has more than 72 publications and was the recipient of the Fitts Human Engineering Award (1997) and the NASA Special Achievement Award (1985). He is also a member of the adjunct faculty of the Wright State University Department of Psychology, where he has taught since 1989. He coedited the volume Principles and Practice of Aviation Psychology with Pamela Tsang.
Pamela S. Tsang is a professor of psychology at Wright State University in Dayton, Ohio. She received her AB from Mount Holyoke College and her PhD from the University of Illinois at Urbana-Champaign. Previously, she was a National Research Council postdoctoral fellow at NASA–Ames Research Center. Her research interests are attention and performance, extra-laboratory-developed expertise, cognitive aging, and aviation psychology. She is interested in applications of her research in a wide variety of domains, including aviation, surface transportation, and medicine. She coedited the volume Principles and Practice of Aviation Psychology with Michael Vidulich.
