Visualization by demonstration

Abstract

A key component of visualization systems that helps human sensemaking is interactivity. Thoughtfully designed interactions make the visual analysis process a conversation between the user and the interface that results in a deeper understanding of data. Yet, despite decades of research, existing visualization systems still require users to interact through layers of menus on control panels. These systems incur extra execution and cognitive cost by introducing a large number of intermediary interface elements such as menus and dialog boxes. This dissertation proposes a novel interaction paradigm for visual data exploration called ``visualization by demonstration''. This paradigm aims to reduce the cognitive cost and enhance interaction expressivity to decrease the level of formalism and fundamental knowledge often required for visual data exploration. This dissertation first discusses the fundamental principles and guidelines that go into the design of visualization by demonstration. It then discusses how we can apply these fundamental principles and guidelines to design and develop general-purpose data visualization tools that implement visualization by demonstration. It finally applies visualization by demonstration to design and develop visual data exploration tools for experts in specific domains such as biology and healthcare.