Shape It Up: An Empirically Grounded Approach for Designing Shape Palettes

Abstract

Shape is commonly used to distinguish between categories in multi-class scatterplots. However, we lack empirically-validated guidance for creating shape palettes that support effective perception. We present a series of three crowdsourced experiments studying how shape characteristics affect category identification in scatterplots. Our experiments measure how shape characteristics including geometry, fill, and orientation affect people's ability to identify point categories quickly and accurately. We find that shape perception is influenced by multiple factors, including a shape's geometric properties and its rendered size. Based on these findings, we derive a set of design guidelines for creating effective shape palettes and demonstrate their application in a proof-of-concept tool for generating shape palettes based on our experimental results.