Proceedings of Louisiana EPSCoR RII Cybertools/Science Drivers 2010 Symposium 1 Toward User Interface Migration for Scientic Visualization Cornelius Toole, Jr. 1, 2 , Brygg Ullmer 1, 2 Kexi Liu 1, 2 , Chris Branton 1, 2 , Rajesh Sankaran 1, 3 1 Computer Science Department, Louisiana State University 2 Center for Computation and Technology, Louisiana State University 3 Electrical and Computer Engineering Department, Louisiana State University Abstract: Visualization is a key component of the scientic investigation process, but dierent contexts (e.g. varying numbers, locales and expertise of users, heterogeneous infrastructure components, etc.) require different toolsets for workow management. We focus on interactive techniques for engaging visualization, especially those that emphasize physical interaction as embodied in tangible and gestural interfaces. In building tangible user interfaces for several software platforms with diverse interaction devices, we face considerable complexity. Here we present preliminary work toward a toolkit with mechanisms for generalizing input from diverse interaction device implementations, as well as generalizing control of a generic visualization model. A key feature of this toolkit is ability to build applications for executing a visualization workow using multiple user interface implementations with minimal code changes. The ability to easily migrate analysis workows between user interfaces moves us toward a future in which users can select the right interactive toolset for a given context. Keywords: human computer interaction, visualization, tangible interaction 1. Introduction Visualization is a key component of the scientic investigation process. And while visualization community has contributed powerful tools for seeking insight from scientic data, no one visualization tool or interaction modality is appropriate for all situations. Currently, the state of the art in user engagement of visualization is fractured. If a developer wants to provide users with software that supports execution of workows that migrate across visualization software platforms, she has to write against multiple architectures. To the extent there are application-programming interfaces (APIs) to common functionality from multiple diverse implementations, developers can use these interfaces to select the capability provider for the current job. We focus on the user interaction aspects of this problem space. One opportunity we see is for the ability to provide common user interfaces to diverse visualization software platforms. Previous work by our group pursued this in the development of a tangible user interface that could drive visualization pipeline instances simultaneously [1]. This work is predicated on the idea that although many visualization applications are based on dierent architectures and internal models, many of them are based on similar general models of visualization (e.g. data ow pipeline or visualization network). Common visualization user interfaces enable users to execute common visualization tasks (e.g. loading data, invoking visualization pipelines for typical lters, changing view parameters, traversing temporal data) across platforms without requiring them to know each of the platform specic menu hierarchies or command languages.