Design and evaluation of the tightly coupled perceptualcognitive tasks in knowledge domain visualization

Knowledge Domain Visualization (KDViz) aims to reveal the dynamics of how knowledge domains evolve over time. It is the ultimate role for information visualization to facilitate a wide variety of user tasks associated with the study of such dynamic, large-scale, transient, and self-organized phenomena. Since the evolution of knowledge domain is necessarily influenced by an array of factors, the design and empirical evaluation of a KDViz system is particularly challenging. In addition, KDViz typically involves a complex and lengthy sequence of data transformation and computational modeling that must be simplified for effective handling by the user interface design. In this article we present a perceptual-cognitive task taxonomy for designing and evaluating KDViz systems. The design implications are illustrated by the development and some refinements of CiteSpace, a Java application that enables users to study how networks of research fronts develop over time. Cognitive tasks such as finding pivotal intellectual points in an evolving knowledge domain are translated into tasks for visual search of objects with visually salient features. The design rationale is explained in terms of the taxonomy, including the use of Gestalt principles, graph drawing aesthetics, and pre-attentive visual search. The implications of the taxonomy on conducting empirical evaluations are illustrated by a case study using evaluation techniques focusing on usability issues. Three levels of user tasks are considered by the taxonomy: 1) the lowest level: manipulating the user interface; 2) the medium level: manipulating the visualization; and 3) the highest level: tracking the growth of a knowledge domain.