Sharon L. Ward

Studies of dynamic task allocation in an aerial search environment

Results of two experiments in dynamic task allocation are discussed. Subjects performed two concurrent computer-based tasks: visual target identification and subcritical compensatory tracking. Target identification could be allocated dynamically to human or computer aid. Three aiding conditions were investigated: no aid, manual aid (with subjects making the allocation decision), and automatic aid (with allocation decisions based on models of human performance). The results indicated that: (1) overall performance was better with the aid available; (2) need for the aid depended on both current and previous task states; (3) unaided performance was benefited by having an aid available, but only if subjects were in charge of task allocation; and (4) although overall performance was better with he automatic aid, subjects preferred the manual aid. The implications of these and other results are discussed. >

An experiment and a model for the human operator in a time-constrained competing-task environment

In a number of supervisory control situations, the human operator is responsible for multiple tasks with different task demands and unclear performance criteria. The operators attention must usually be allocated between a number of competing systems. Quite often, due to the dynamic nature of the system, quick actions may be required at various points in time. To study the human in such a time-constrained environment, an experiment was designed based on monitoring the spread of forest fires to identify threatening situations. Results from the experiment were used to identify heuristics which, in turn, were used in a model of the human operator. The model calculated a set of feasible control actions at any time, from which a single action was chosen based on priority indices calculated using fuzzy membership functions. The model matched observed results very well.

Human Attention Allocation Strategies in Situations with Competing Criteria - Development of an Experimental Paradigm and Outlines for a Model

Abstract In supervisory control situations involving multiple human operators, proper cooperation and coordination among the operators is essential. In addition to their individual responsibilities, the operators are jointly responsible for certain aspects of the system. Time available for decisions may be limited, and various criteria may conflict with each other. Proper scheduling of tasks is necessary for optimal performance. In such situations it is usually not clear when and who should take responsibility for the joint tasks. Understanding how multiple operators interact requires the understanding of single operator performance. A model based on Pareto optimality and fuzzy set theory has been proposed for the human operator. An experimental paradigm has been developed based on a threat assessment situation. Features of the paradigm are explained. A detailed outline for the model structure is given, and experimental results are discussed.

Experimental Evaluation of Adaptive Task Allocation in an Aerial Search Environment

Abstract A classic issue in the study of man-machine systems concerns the allocation of functions or tasks among humans and machines. The traditional static approach to task allocation, which rests on the premise that humans perform some types of task better than computers and vice versa, is unsatisfactory for a variety of reasons. It has been suggested that a dynamic, adaptive approach to task allocation, possible with the increasing sophistication of computer technology, shows potential for improving system performance Several issues which should be considered when designing such an adaptive aid are being investigated empirically within the context of an aerial search environment. This environment is described and available experimental results are presented. The results indicate that overall system performance may be improved by adaptively allocating tasks to human or computer, and suggest that the availability of an adaptive aid may offer potential benefits to unaided human performance as well

EXPERIMENTAL EVALUATION OF ADAPTIVE TASK ALLOCATION IN AN AERIAL SEARCH ENVIRONMENT

A classic issue in the study of man-machine systems concerns the allocation of functions or tasks among humans and machines. The traditional static approach to task allocation, which rests on the premise that humans perform some types of task better than computers and vice versa, is unsatisfactory for a variety of reasons. It has been suggested that a dynamic, adaptive approach to task allocation, possible with the increasing sophistication of computer technology, shows potential for improving system performance. Several issues which should be considered when designing such an adaptive aid are being investigated empirically within the context of an aerial search environment. This environment is described and available experimental results are presented. The results indicate that overall system performance may be improved by adaptively allocating tasks to human or computer, and suggest that the availability of an adaptive aid may offer potential benefits to unaided human performance as well.

The Use of Brain Evoked Response in a Critical Event Detection Task

Brain evoked potentials (EP) were used to study subject responses to “critical events” (CEs) which were part of a series of stimuli. Both reaction times (RT) and EPs were influenced by the appearance of the CE. Longer RTs and a large late positive component of the EP were associated with the trials containing the CE stimuli. Intermediate RTs and late positive components of the EPs were found in a condition in which the stimulus preceding the CE gave information about the following CE and was also similar in appearance. While this is a laboratory test, it is similar in many ways to actual situations confronted by human factors specialists. EPs appear, then, to be useful tools in the system design and evaluation process.

Changes in Maze-Solving Errors Due to Stress

Maze-solving was examined as a candidate performance measurement tool for assessing the effect of stress. Stress was induced by limiting the amount of time available for subject input. The usefulness of this task as a measurement tool was evaluated in a series of three experiments. The first experiment sought to isolate a particular maze configuration (size and difficulty level) which would show reliable and sensitive differences among a range of dot speeds. The second experiment assessed the effect of repeated presentation on performance. The final experiment in this series investigated the effect of dot speed on maze performance using a within-subject design and the maze configuration chosen in the first experiment. Results from these experiments were consistent: increases in dot speed produced reliable decrements in performance. The strategies used in solving these mazes were examined qualitatively as well as quantitatively. The number of incorrect paths chosen increased as the dot speed increased, and errors that did occur were not as readily corrected. Furthermore, the errors that occured with increased dot speed were typically errors of commission, rather than simple failures of motor coordination. These tendencies indicated that the decrement in performance scores with increased dot speed was due to shortened planning horizon.

Information weighting and combination in the combat supervisor's tactical decision behavior

Discriminant analysis procedures were used to describe mode selection in a simulated AAA system. Mode selection variability and shifts in selection strategies were found across conditions. These differences were attributed to decision dependent momentary shifts in subjective information weighting.

Application of Preference Tree Methodology to Choice Behavior in a AAA Supervisory Control Task

An attempt was made to extend a model of preference, known as elimination-by-aspects (Tversky, 1972), to situations involving active search for sequentially available information of varying value. Subjects were instructed to choose the most appropriate one of three possible operating modes in a task simulating supervisory decisions in a AAA system. The data failed to support the EBA model. Instead of searching aspects associated with each alternative in a non-compensatory fashion, as predicted by the model, subjects searched the information set in a compensatory manner. Failure of the model is discussed in terms of three points (a) subject control of information search, (b) sequential information availability, and (c) variability in the value of information across conditions. Consideration of these three points leads to the conclusion that the usefulness of the EBA approach is limited to situations in which the subject has all relevant information available to him before consideration of alternatives.