SSC San Diego TR 1864, June 2001

Approved for public release; distribution is unlimited. ACKNOWLEDGMENTS The author is grateful to have contributed to workload analysis and numerous other aspects of the Multi-Modal Watchstation demonstration project. Specific thanks to Christopher Griffith for software development that allowed capture of workload-related indices, and to Kimberly Heinlein for assistance in obtaining workload data during usability testing. Assessing and understanding operator workload is an important factor for consideration during the development of new systems. It may also be important to understand fluctuations in workload within operational systems in order to efficiently apply automated processes and provide assistance at critical times. This paper describes how a simple workload measure obtained every 2-3 minutes during the evaluation of a prototype command and control console can be used to develop an operator's workload profile as a function of other system parameters, such as track density on the tactical plot, and task loading. If system measures can be monitored regularly, functional models of operator workload can be derived, and workload levels can be interpolated to provide near-continuous work-load estimates every 15-30 seconds. The resulting workload profiles can be used to identify conditions that result in potential operator overload. Profiles from several operators may be used to study team workload distribution and to derive more efficient work allocation strategies. Also discussed in the present work is how the simple unidimensional workload measure relates to multidimensional measures that differentiate between mental demand, physical demand, frustration, and other aspects of work. Multidimensional measures require more extensive reporting and are thus not suitable for administration during system testing. A common multidimensional scale, the NASA Task Load Index (TLX), was administered at the conclusion of the each evaluation session. Regression analysis revealed that the 90th percentile from the distribution of unidimensional workload estimates related to the NASA-TLX dimensions of mental effort and temporal demand in a group of 20 operators. These findings indicate that near-continuous workload profiles may be built from simple subjective workload estimates combined with system-state information, and that the workload estimates can be linked to specific behavioral dimensions as captured by more complex workload assessment scales. 4. Mean task load (left axis, solid line) and mean subjective workload (right axis, open circles) as a function of time (in sec) for the eight subjects. Mean task load was obtained every 15 sec. Mean subjective estimates were forward-lagged by 60 sec.... 5 5. Workload estimates produced every 15 sec by …