Effect of Test Result Uncertainty on the Performance of a Context-Free Troubleshooting Task.
Abstract
A major problem affecting weapon system support cost and operational capability is the 20-40% troubleshooting error rate experienced by the Air Force and the Navy for all aircraft corrective maintenance actions. Most fault detection problems resemble the problem of signal detection in a background of noise, or in the case of troubleshooting, test result uncertainty. This thesis attempted to determine whether human bias exists under conditions of test result uncertainty such that troubleshooting performance is systematically affected. A knowledge of such bias might be useful in assessing the utility of powerful signal detection-in-noise analytical tools, e.g., the Relative Operating Characteristic (ROC) curve analysis, to improve predictions of troubleshooting performance, and reduce troubleshooting error by allowing man-machine troubleshooting systems to be optimized on the basis of the response characteristics of both machine and man. The experiment consisted of sixty-four subjects performing a simulated electronics troubleshooting task at four levels of test result error; zero error, 25% error, 50% good called bad error, and 50% bad called good error. Results did not indicate the existence of sognificant differences in troubleshooting performance under the latter three treatment conditions; only between the first and any of the latter three. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 17, 1982
- Accession Number
- ADA128584
Entities
People
- Harry A. Baran
Organizations
- Air Force Institute of Technology