Theoretical Foundations of Information Visualization - Information Visualization: Human-Centered Issues and Perspectives - page 60

Original Visualization Session Graph Refined Visualization

Fig. 2. Analysis and evolution of a network traffic visualization. The original interface (a) uses
colored lines connected to the edges of a square to depict changes. A formal model was used to
capture interaction with the tool, and these sessions were analyzed to improve the interface (b).
The redesigned interface (c) makes the exploration more efficient by displaying all event types
individually and combined.

There are several benefits to a complete visualization exploration model. An under-
standing of how humans process visual cues in order to make exploration decisions
can inform visualization design. For example, this “information scent” has been used
to understand the cost-benefit trade-offs of different focus+context visualizations and
to formally understand efficient web interfaces [31,32]. Similarly, a formalism for the
visualization process lends itself to analysis to measure the user's efficiency of explo-
ration [33,34]. Clusters of similar results (based upon metrics) during the session
suggest redundant exploration; analysis of sessions based upon these metrics illumi-
nate the path to more effective design [29,35] (Figure 2).

4.4 Visualization Transform Design Model

An exploration model describes and predicts a human's interaction with a visualiza-
tion system based upon its design. This model neglects to describe the components
that compose the design or provide initial design guidance. To provide this guidance,
visualization transform design models are needed. A visualization transform is the func-
tion that computes the depicted result from visualization parameters - elements such as
brushed graph nodes or opacity maps that dictate the rendered result. Significant work
has expressed different mechanisms for constructing such transforms [36,37,38,39,40]
(see Figure 3 for an extended example), but these categorizing efforts lack two things
to provide a formal foundation. First, they do not utilize perceptual and cognitive
literature to suggest and evaluate design decisions; second, most do not address the
evolution of transforms and the utilization of extra-visualization tools important to the
analysis (e.g., statistical packages). Efforts in providing design guidance for data
display has been investigated [41,42,43] and formal algebras for transform modifica-
tion have been recently presented [44,45]. However, these distinct contributions re-
quire effort to unify and to validate for a complete and cohesive scientific foundation.
A complete, predictive transform design model will yield several benefits. Toolkits
for visualization creation benefit by providing guidance on suitable, less suitable, and
unsuitable component choices. Visualization pedagogy will improve due to a vali-
dated foundation for techniques. Further, formal models will lead to objective metrics
for evaluating a transform's effectiveness. All of these enhancements feed back into a
visualization system, improving its potential utilization.