A Comparison of Three Methods of Task Analysis: Cognitive Analysis, Graph-Matrix Analysis, and Self-Organizing Networks

Abstract

Three methods of performing a task analysis are compared: cognitive analysis, graph-matrix analysis, and self-organizing networks. Cognitive analysis relies on the ability of an observer to abstract and generalize over situations. Graph-matrix analysis is valuable for its precision an inclusion of details. Neural networks have an ability to generalize uninfluenced by observer bias. Comparison demonstrates that each method misses some important but different aspect of human-computer interaction. The cognitive analysis included infrequently used information that was not captured by direct observation. The graph-matrix analysis included frequency of use information and details missed by the cognitive analysis. The self-organizing network generated an alternative view of the task structure that was not influenced by observer bias. It showed that the underlying structure for the user-computer interaction in this study was the structure of the computer system itself.