limited the generating and selecting functions to the human, whereas, Decision Support permitted both human and computer responsibility for task planning. This difference in the function allocation schemes explains the observed difference between the two LOAs in terms of time-to-recovery in the simulation. Batch Processing involved subjects in the simulation to a greater extent than Decision Support possibly promoting heightened states of subject system awareness for efficient recovery.
In summary, these results are supportive of lower level automation (Batch Processing) maintaining human involvement in the control loop during normal system functioning. This reduced the time-to-recovery during failure modes.
ANOVA results on the average percent correct responses to SAGAT queries covering the three levels of SA revealed a significant effect of LOA (F(4,9) = 3.4, p < 0.05) only on Level 3 (system state projection). The mean percentage of correct responses to Level 3 SA queries decreased from low to intermediate LOA and increased slightly from intermediate to higher LOAs. According to Tukey's HSD test, Action Support was significantly superior to Batch Processing in terms of the percentage of correct responses. Action Support varied from Batch Processing in that the latter stripped subjects of the capability to manually control the robot using the SpaceBall®. This difference between the two LOAs was associated with a degradation in, for example, operator ability to predict the next move sequence in the telerobot task.
ANOVA results revealed LOA to be significant in its effect on the NASA-TLX, F(4,9) = 22.538, p < 0.001. The general trend of perceived workload decreased at progressively higher LOAs. Tukey's HSD test on the overall workload scores revealed Action Support to significantly differ (p < 0.05) from all other levels and the same to be true for Batch Processing. This result reflects the reduction of human involvement in the active system control loop. As subjects were progressively removed from plan implementation (Batch Processing), limited in their capability to plan move sequences (Decision Support), and reduced to the status of system monitor (Supervisory Control) or observer (Full Automation), workload significantly decreased.
This study reviewed LOA taxonomies presented in the literature. Arguments both for and against the use of general taxonomies were levied. In particular, a potential lack of applicability of many LOA taxonomies to real-world tasks was raised. This issue was addressed through further empirical exploration of Endsley and Kaber's (in press) LOA taxonomy demonstrating the applicability of specific LOAs to a simulated telerobot task. The study demonstrated that higher LOAs enhance performance during normal operating conditions through computer processing. This was accompanied by lower levels of subjective workload observed at the same levels. The experiment revealed intermediate LOAs to promote higher operator SA and enhance human manual performance during system failure modes as compared to high level automation. This effect was also attributed to maintaining operator involvement in the system control loop during normal operations.
From an automation design perspective, this research validates LOA as an alternate approach to traditional automation. It provides detailed guidance to teleoperation/telerobotic systems designers in allocating system responsibility to a human and computer for safe performance in hazardous operations.
Endsley M. R. ( 1988). "Design and evaluation for situation awareness enhancement". In Proceedings of the Human Factors Society 32nd Annual Meeting (pp. 97-101). Santa Monica, California: Human Factors and Ergonomics Society.
Endsley M. R. & Kaber D. B. (in press). "Level of automation effects on performance, situation awareness and workload in a dynamic control task". Ergonomics.
Endsley M. R. & Kiris E. O. ( 1995). "The out-of-the-loop performance problem and level of control in"