Interactive techniques for exploring data using behavioral animation

Data driven behavioral animation is a new visual computing technique that involves mapping data attributes to the behavioral functions of a set of graphical data avatars. When the avatars are allowed to interact their individual and group behavior reveals qualitative structures in the data as well as relationships. In this paper we present a set of novel techniques for interacting with systems of data avatars. These are especially intended to support the user in exploring the data visually in the early stages of data analysis. We show that the methodology supports a comprehensive set of interactive tasks including the abstracting of meaning, comparing groups, identifying relationships, imposing structures and hypothesis testing. The techniques are implemented in a system called AnimVis that supports behavioral data exploration in a two-dimensional arena.