Cagatay Demiralp, Carlos E. Scheidegger, Gordon Kindlmann, David Laidlaw, Jeffrey Heer. IEEE Computer Graphics and Applications’ Visualization Viewpoints, 2014.
In this paper, we present an operational definition of “good visualization” based on matching mathematical structures of the input and output spaces. We introduce visual embedding as a model for visualization construction. We define a visualization as a function that maps from a domain of data points to a range of visual primitives. We claim a visualization is good if the embedded visual elements preserve structures present in the data domain. Concretely, we show how striving to preserve metrics leads to implementations of colormap design, glyph selection and quantitative evaluation algorithms.
Get the paper here.