Bad for Data, Good for the Brain : Knowledge-First Axioms For Visualization Design

Abstract

Traditionally, visualization design assumes that the effectiveness of visualizations is based on how much, and how clearly, data are presented. We argue that visualization requires a more nuanced perspective. Data are not ends in themselves, but means to an end (such as generating knowledge or assisting in decision-making). Focusing on the presentation of data per se can result in situations where these higher goals are ignored. This is especially the case for situations where cognitive or perceptual biases make the presentation of “just” the data as misleading as willful distortion. We argue that we need to de-sanctify data, and occasionally promote designs which distort or obscure data in service of understanding. We discuss examples of beneficial embellishment, distortion, and obfuscation in visualization, and argue that these examples are representative of a wider class of techniques for going beyond simplistic presentations of data.