Out‐of‐the‐loop performance problems and the use of intermediate levels of automation for improved control system functioning and safety

Abstract

Human supervisory control and monitoring of automated systems, as well as, passive system(s) information processing can all be classified as forms of out‐of‐the‐loop (OOTL) performance. Whether the operator's task is to decide if process control intervention is necessary, detect a critical system event, or accept or reject the actions of a computer controller, he or she is removed from direct, real‐time control of the system. OOTL performance is a critical issue in overall automated systems functioning because it is associated with numerous negative consequences including: (a) operator failure to observe system parameter changes and intervene when necessary (vigilance decrements); (b) human over‐trust in computer controllers (complacency); (c) operator loss of system or situation awareness; and (d) operator direct/manual control skill decay. These consequences have been found to impact human performance under both normal operating conditions and system failure modes, with a greater effect on the latter [15] leading to serious problems in operator ability to perform their assigned tasks when working with automated systems. Level of automation (LOA) has been put forth as an approach to ameliorating OOTL performance problems. It is intended to determine the optimal assignment of control between a human operator and computer in order to keep both involved in system operations. LOA considers the capabilities and capacities of both the human and computer controller in determining their optimal coupling. It constitutes a systems approach to resolving OOTL performance problems by minimizing the negative consequences associated with the removal of the operator from active system control, and allows for the strengths of both human decision making and computer processing to be realized. When compared to a technological approach that assesses only the capabilities of the computer in allocating as much responsibility to the machine as possible, and assigning the remaining tasks to the human operator, the advantages can be considerable. A LOA taxonomy will be presented along with research examining its utility in a dynamic control task. Using LOA to identify optimal combinations of human and computer control was found to produce improvements in system performance under intermediate levels. These levels involve joint human and computer control of various system functions, such as monitoring, planning, and option selection and implementation. Results indicated decreases in the number of system processes/tasks overlooked by operators. These improvements may translate into cost reductions due to improved operational safety and are anticipated to be applicable to process control operations.