|Résumé||It’s well known that visualizations can lie, that is, distort the true relationship between the variables and records being represented visually. Example precursors of this topic were Huff and Geis How to Lie with Statistics, Tufte with indirectly measuring the synthesis of the visualization's chart noise and ink metrics, and Bertini, Ward, and others with visualization quality metrics. The more direct question we would like to answer is how to tell if a visualization is accurately representing its data. Or more precisely how can we measure the truth in a visualization.
In this presentation we will describe three different approaches to measuring the truth in visualizations. First we will define what we mean by truth. We then will describe Weave, an open source analysis and visualization environment [iWeave.org] and our architecture for computing descriptions of visualizations developed for ADA compliance within Weave. Finally we will give preliminary examples of the measures.
This research work has broad implications and applications to applied visualization. |