Display Formatting: An Expert System Application

Abstract

Expert systems must often rely on probabilistic information to solve complex problems. Computer output may consist of more than one possible solution, leaving the final decision to the user. The amount of output and its mode of presentation need to be tailored to the application. In situations requiring fast and accurate decision making, as in military systems, output must be presented in a concise and easy to interpret format. Six presentation variables which could affect alternative selection performance were examined in an experiment: information quantity (4 or 8 numbered alternative values), format (tabular or bar graph), order (ordinal presentation of the alternative numbers or their associated values), bar graph layout (vertical or horizontal bars), bar graph labels (no labeling or labeling of each bar with its associated value), and table type (probability or rank values). The variables were combined into 24 different display formats; each of 24 subjects was tested on all the formats. The task consisted of finding the probability or rank values in the order specified on the top of the CRT screen (high value to low or low to high) and entering the associated alternative numbers on a keyboard. Mean response time and mean errors, over five trials, were the measures of performance on each display format. Results of the analyses indicated that the best display format in terms of the speed and accuracy of selection performance was the 4-alternative vertical bar graph with ordered probabilities and labeled bar values.