Adaptive Aiding for Human-Computer Control: Experimental Studies of Dynamic Task Allocation.
Abstract
Adaptive aids are those that allocate, partition, or transform tasks dynamically in response to system or operator state in order to maximize system performance. The ultimate goal of this research is the identification of guidelines for the implementation of adaptive aids which can be useful to system designers. In the experimental task environment developed for this research, subjects perform a subcritical compensatory tracking task while simultaneously identifying targets (spotting) on a graphic display that moves down a CRT screen. A computer aid capable of identifying targets is sometimes available to perform the spotting task. The aid and spotting task are designed such that the relative superiority of human and computer may be expected to change over time; hence the spotting task should be allocated dynamically to human or computer for best overall performance. The results of two experiments on dynamic task allocation are presented in this report. In the first experiment, subjects performed both tasks with and without the spotting aid under various levels of tracking difficulty. Activation of the spotting aid was totally under subjects' control, and they were free to use the aid whenever they wished. Based on the results of this experiment, multiple regression models predicting subjects' performance in various task conditions were developed. Keywords: adaptive expert systems.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1986
- Accession Number
- ADA166704
Entities
People
- Nancy M. Morris
- William B. Rouse