Structural analysis of verbal data

Abstract

Current methods of analyzing verbal reports (Protocol Analysis) from human-computer interactions fall short of their potential. Although there are systematic methods for collecting complete and objective verbal reports applicable to a broad range of problem-solving tasks, currently available analyses of verbal reports are ad hoc and apply only to well constrained tasks. Structural Analysis is a systematic method, currently under development, for analyzing real-world tasks involving human-computer interaction. Starting with a rule that assigns utterances to two dichotomous categories related to a behavior of interest, rules are generated that expose the goal building and evaluation underlying that behavior. The resulting data yield time distributions that characterize subjects' goal-directed behavior and that allow comparisons among tasks or among subjects.