 |
|
|
Background Human error is the causal or contributing factor in the majority of aviation accidents. It is now well understood that these errors are the product of both the propensity of human to make errors and specific design features that may induce errors. Attempts to reduce human error by improved design have met with only partial success. A significant obstacle to error tolerant design is the complexity of human behavior itself. Simple tests of system interfaces are insufficient to uncover the wide range of errors human operators will make in using them. The nature of human error in aircraft accidents is often puzzling since the crew typically will have performed the same sequence of tasks many times. Thus the accident data shows a seemingly capricious tendency for error, making the measurement of the benefit of new error tolerant systems problematic. For the same reasons, simulations often fail to provide adequate estimates of human error, yet simulations remain the most useful means of examining human behavior in circumstances approximating operational condition. Complexity affects simulations in an additional way: it is difficult to measure and quantify important aspects of human behavior because of the variability inherent in complex performance. Improved understanding of how errors are generated in performing tasks would facilitate the design of error tolerant (or error resistant) systems. Improvements in measuring complex performance will be necessary to adequately test new designs. Objectives The proposed work seeks three objectives:
Approach Instances of error often reflect failures of executive control that result from limitations in human information processing. Failures of executive control lead to identifiable classes of memory errors, such as failures to remember intended actions (prospective memory failures), and habit capture error. A review of ASRS incidents revealed a significant number of such memory failures. Failures of executive control are also associated with failures in routine monitoring, where observers will fixate an information source but not apply the executive control needed to process the information. Examination of NTSB accident reports shows that in almost all cases, failure to note obvious discrepancies or the failure of the pilot not flying to perform cross-checks are cited as causal or contributing factors in the accident. In current theories of human cognition, executive control is associated with limited-capacity attention-demanding mental processing. Common cognitive acts such as fetching items from memory, reading text, solving problems, etc., require executive control, in contrast to early perceptual processes and low-level motor behaviors whose processing is independent of executive control. Common occurrences, such as the familiar situation of going through the motions of reading a passage without comprehension, reflect the dissociation between executive control and automated processes (in this case motor programs). A better understanding of the relationship between executive control and the more autonomous information gathering and motor behaviors would lead to significant advances in our understanding of human error. One pressing question is the identification of specific task actions that require attention. The abstract stage models developed through empirical experimentation can be elaborated through imaging techniques that record brain activity directly. The cognitive variables of most interest are not directly observable. This has made it difficult to relate theory to complex applications domains where measurement techniques are too crude for the required inferences. Unobtrusive, but precise, measurement techniques like eye fixation recordings and brain imaging promise to provide improved understanding of the mental demands of complex tasks such as piloting. The feasibility of these techniques should be examined. Sub-Tasks Sub-Task 2-1: Eye-Movement Metrics for Human Cognitive Analysis and Modeling Sub-Task 2-2: Models & Metrics of Human Executive Control Sub-Task 2-3: Models & Metrics of Human Spatial Attention and Memory Point of Contact
|
|
|