Publication Date
2019
Document Type
Dissertation
Committee Members
Joseph Houpt (Advisor), Pamela Tsang (Committee Member), Assaf Harel (Committee Member), Gregory Funke (Committee Member), Tamera Schneider (Committee Member)
Degree Name
Doctor of Philosophy (PhD)
Abstract
The progression of technology and adaptive automation has shown tremendous promise in reducing both physical and mental task demands, while allowing the maintenance or improvement of overall performance. Consequently, a user is able to maintain task performance with relatively more ease and reallocate spare time and energy to additional tasks. Quantifying the resources that one has left is an ongoing, relatively open, research objective for human factors psychologists. Here, we created a standardized, individual-level metric to serve as an estimate of multi-tasking efficiency. We go beyond just rank-order, or categorical labels of suffering from, benefiting from, or adequately maintaining performance in multiple tasks. We quantify the extent that participants could sufficiently take on additional tasks by comparing their observed performance to what their performance would be assuming they were an efficient multi-tasker. A multi-task context necessarily has more task demands and assuming that each single-task is relatively challenging in isolation, is predicted to cause performance decrements. The degree that performance decrements occur may vary depending on the characteristics of the single-tasks. Specifically, multiple resource theory (MRT; Wickens, 1984) provides a theoretical modeling framework to make predictions about what tasks will produce more (or less) performance decrements when paired together. In Chapter 2 we used a computational characterization of MRT to compute multi-task conflict in a three dual-task subsets of a larger relatively complex multi-tasking environment. We used the MRT measure of conflict to make rank-order hypotheses about which specific dual-task pairs should cause more (or less) performance decrements relative to the others. In Chapter 3 we develop a measure of multi-tasking throughput (MT) and test MRT hypotheses by computing and comparing participant’s MT coefficients across several multi-tasking combinations. In Chapter 4 we argue that increasing the temporal precision of MT will extent it’s usefulness beyond offline analyses and into real-time estimation of multi-tasking performance, at the individual-level. To capture the fluctuations in MT over time we developed a Bayesian model of MT to estimate performance efficiency for smaller segments of time. We found participants’ performance statistically significantly varied across time, and the pattern of change was unique across individuals and for particular multi-task conditions. In Chapter 5 we used a machine-learning approach, specifically support vector machines, to predict the class of neural activation in two demand manipulations. We tested the accuracy of each within-subjects model to predict particular task demands, given one’s neural activity, in the same-day and in a different day. We compared the accuracy of several subsets of different electrodes and bandwidths pairs, to a set that contained all pairs. We found across individuals, task demand manipulations, and days that the model provided the full set of electrode and bandwidth pairs outperformed all subsets. In Chapter 6 we wanted to further investigate whether there was a pattern by which the machine classification model was making mistakes, i.e., when it misclassified a particular neural data point. Specifically, we were interested in whether the pattern of misclassification was correlated with the pattern of structural resource competition (MRT) we quantified in Chapter 2, and/or performance conflict (MT) that we quantified in Chapter 3. Our data did not support our hypotheses that the degree of separation between each single task condition in a neural MDS representation would correlate with resource competition (MRT predictions) and dual-task MT. Further, we did not find that those who exhibited more performance deficits, also exhibited more neural similarity in those conditions relative to others. In multiple Chapters we tailored a discussion that situated our findings relative to the field and provided a plethora of fruitful routes to take in future research endeavors.
Page Count
194
Department or Program
Department of Psychology
Year Degree Awarded
2019
Copyright
Copyright 2019, all rights reserved. My ETD will be available under the "Fair Use" terms of copyright law.
