A Desktop Expert System as a Human Factors Work Aid

The advent of increasingly powerful microcomputers, coupled with the development of small, feature-packed expert systems now makes it cost effective to provide workers with relatively inexpensive desktop expert systems. In order to evaluate the value of such systems as work aids for human factors engineers, we developed a small demonstration system using a commercially available expert system development tool, NEXPERT^TM, released in 1985 by Neuron Data, Inc. of Palo Alto, CA. We selected a candidate problem area based on four criteria: 1) the problem domain had to be small enough to be covered comprehensively by a relatively small knowledge base; 2) the problem domain had to be potentially useful to video display terminal (VDT) screen designers; 3) appropriate information had to be readily available in human factors guidelines, published reports, and journal articles; and 4) the problem should provide the opportunity to exercise as many of the features of NEXPERT as possible. The topic area we selected was “video display screen color”. Our goal was to produce a job performance aid (JPA) that non-human factors VDT screen designers could use to select appropriate colors for screen features. Because the system users typically have little or no formal training in human factors, the JPA has to supply color recommendations in the form of clearly stated requirements, but with the decision rationale and additional references also immediately available for users wanting more information. Using the expert system shell provided by NEXPERT, we constructed a knowledge base containing more than one hundred IF …, THEN … rules representing knowledge gained from a detailed literature review. We initially validated our expert system by posing a wide variety of hypothetical design problems and assessing its conclusions against our expectations. Based on our work so far, we have concluded that small expert systems can be useful in providing human factors expertise to system designers. We believe that increasing use of expert systems may soon lead to changes in the typical current scientific publication format to include knowledge base rules provided by the author(s).