The Effects of Levels of Invocation Authority on Adaptive Automation of Various Stages of Human Information Processing
No Thumbnail Available
Files
Date
2003-02-09
Authors
Journal Title
Series/Report No.
Journal ISSN
Volume Title
Publisher
Abstract
Adaptive automation (AA) has been explored as a solution to problems associated with extended periods of manual control and static automation in complex systems including fatigue, high workload, and loss of situation awareness. Several important decisions must be made regarding the implementation of AA in the design process, including who will decide when to invoke automation and what function or functions will be automated. Several studies have recently proposed models of automation based on four stages of human-machine system information processing, including information acquisition, information analysis, decision making, and action implementation. It is possible to apply AA to several of these functions simultaneously or independently. Management authority over automation (invocation authority) can be extended to a human operator, to a computer controller, or distributed between both through mutual suggestions and approvals of each authority. Both invocation authority over dynamic function allocations (DFAs) and the type of AA have both been shown to independently influence task performance, situation awareness, and workload in adaptive systems. It is possible that different levels of authority may have different effects on overall humanmachine system performance when applied to various automated functions, such as detection tasks as compared to tasks requiring higher cognitive functions, like decision making.
The goal of this study was to assess the performance and workload effects of applying AA to four stages of human- machine system information processing using a performance-based approach and facilitating DFAs through two levels of computer authority (suggestion and mandate). The research was expected to provide insight into the existence of any interaction between these aspects of AA design. It was expected that higher level automation, such as information analysis and decision making, would be more compatible with mandated allocations, while lower levels, such as information acquisition and action implementation, would be more effective under partial human control. The additional task load imposed by the requirement of operator acceptance of DFAs suggested by a computer authority was expected to adversely affect performance of the more cognitive tasks.
Forty naïve subjects performed an air traffic control task as part of a dual-task scenario in which a secondary, gauge- monitoring task served as the objective measure of workload that controlled DFAs in the primary task. Each subject experienced one of four forms of automation (or no automation for control subjects) and both types of authority (suggest and mandate) during two trials (Each trial incorporated only one of the two authority types.). Results confirmed performance differences due to AA across the various aspects of information processing as well as between the two types of invocation authority. Specifically, subjects performed significantly better in the primary task during periods of automation as part of information acquisition AA as compared to decision making. During those same automated periods, subjects also performed significantly better when automation was suggested as compared to mandated. Contrary to the central thesis of the work that there would be a negative performance effect when computer suggestions were combined with AA of higher cognitive functions (information analysis and decision making), there was no evidence of an interaction effect of the two experimental manipulations. However, the individual results regarding automation type and authority were consistent with results appearing in the literature.
The results of this study provide evidence that the effectiveness of AA is dependent upon both the type of automation presented to an operator and the type of invocation authority designed into the system. This could potentially provide additional insight for effective AA design in complex systems.
Description
Keywords
automation, adaptive automation, human information processing, cognitive ergonomics
Citation
Degree
MS
Discipline
Psychology
