Effective Interactions for Personalizing Spatial Visualizations of Collections

Kenneth C. Arnold, Krzysztof Z. Gajos

Abstract: Interactive spatial visualizations powered by machine learning will help us explore and understand large collections in meaningful ways, but little is yet known about the design space of interactions. We ran a pilot user study to compare two different interaction techniques: a "grouping?" interaction adapted from interactive clustering, and an existing "positioning?" interaction. We identified three important dimensions of the interaction design space that inform future design of more intuitive and expressive interactions.