| Numerous catastrophic accidents have been the result of human operators making poor judgement calls stemming from suboptimal decision making. This suboptimal decision making, in many cases, arises when an operator is either in a high cognitive workload state, overwhelmed with information leading to a greater chance of missing an important detail, or in a low cognitive work- load state, distracted and overall not paying attention to the task at hand. If the cognitive workload of an individual can be properly monitored, suboptimal operator conditions can be recognized or prevented, reducing the chance of an accident. While previous research has led to specialized, single subject classifiers of cognitive workload, little work has been done in the way of generalizing the classification model across multiple people, which would reduce the amount of training needed and provide insight into underlying patterns. This thesis explores the effectiveness of a cognitive workload model under conditions in which the test participant is not included in the training dataset. Additionally, two loading tasks are used to evaluate how well a model trained on one task can perform on another task. The results gathered support the idea that a generalized cognitive workload model that can be utilized with individuals and tasks not included in the training dataset is possible. This novel classification model opens up the possibility for future research into the similarities and differences of individuals with regards to cognitive workload. Additionally, if the model can be expanded upon to increase the classification accuracy, a system can be developed that monitors the operator in real time to provide warnings or modulate automation. |
