A state transition analysis of online information-seeking behavior

Statistical analyses of online searching patterns compared the usage of a query language by various groups of searchers. Data were gathered by an experimental project, Individualized Instruction for Data Access (IIDA), concerned with developing and testing a system which serves as a teacher and assistant to users of Lockheed's DIALOG system. Sequential listings of user commands were classified by corresponding state codes to represent phases of searching. Zero- through fourth-order Markovian analyses of individual commands and strings of like commands were performed to compare searching procedures used by three classes of users. Class 1 compared novices differing in their use of IIDA as an adjunct trainer; Class 2 compared inexperienced, IIDA-trained searchers with professional searchers; Class 3 compared inexperienced searchers taught by either IIDA or professional searchers. Analysis revealed that searchers tend to follow the advice of their trainer, whether it be human or computer. The use of IIDA as the sole training method appears to produce searchers who use similar patterns of searching. The use of another training method in addition to or instead of IIDA appears to produce searchers with varied and individual “styles” of online searching.