Human Problem Solving in Fault Diagnosis Tasks

Abstract

Abstract : This report evaluates the nature of human problem solving abilities specifically related to fault diagnosis situations. Three types of fault diagnosis were analysed involving diagnosis utilizing both real equipment and computer simulated equipment failures. In addition, the investigators experimented with computer generated problem solving aids to supplement human decision making capacities in diagnostic tasks. Results of the initial investigation indicate that human problem solving tends to be highly context- specific but that pattern recognition capabilities are exceptional allowing for a high degree of accuracy in ambiguous problem solving situations. Both structured-oriented and strategy-oriented problem solving aids were analysed. Structured-oriented bookkeeping aids clearly improved performance, while strategy-oriented aids actually had a negative effect on transfer of training. This research effort is clearly relevant to military interests in effective training of optimal decision making in sub-optimal conditions. The results strongly support further research into the nature and training of effective solving. Keywords: Troubleshooting; Fuzzy set model; Complexity model; Rule- based model; Training.