Towards a Role of Visualization in Social Modeling

The traditional role of visualization in large scale social modeling projects has mostly been relegated to presentation and reporting. While these projects see the potential of communicating information visually, the visualization component has too often been considered as the final stage of a long process, sometimes as a lastminute add-on. The result is that these visualization components are limited in capabilities, and often appear disjointed and forced from the rest of the project. In this paper, we propose that for visualization to be more commonly accepted, it needs to fit the role as an analytical tool on top of being a presentation component. We use our probe-based visualization as an example of how such cross-over can occur, and present some challenges in social modeling that can be addressed using visualization techniques. Visualization is becoming prevalent in our everyday lives. Popular websites such as Google Maps, the Visual Thesaurus, and Baby Name Voyager have increased the public’s awareness of how information can be efficiently communicated visually. In the past few years alone, visualization components are appearing with more frequency in the final products of large scale social modeling projects. However, a gap still remains between how the visualization community identifies itself and the perception of the role of visualization by the other disciplines. For too often visualization is considered to be the last step of a project, where visualization experts are asked to “make pretty pictures out of the final data” without having participated in the design or implementation of the overall system. We propose that for visualization to take a more integral part in these systems, it needs to fulfill both the roles of a presentation tool as well as an analytical tool. Not only does visualization need to deliver the “pretty pictures,” it also needs to provide the analytical capability that is difficult for non-visual components to deliver. As an example, we describe our probe-based visualization that supports interactive analysis and presentation of an agent-based simulation. We then propose additional challenges in social modeling that can be addressed by integrating a visualization component. Copyright c © 2009, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. Visualization for Presentation and Analysis We designed a probe-based visualization as an interface to an agent-based model for simulating the social-economic factors of Afghanistan (Butkiewicz et al. 2008). Although the interface is designed independently of the simulation, it achieves both goals of presentation and analysis. On the presentation level, the interface allows the user to interactively select regions-of-interest and presents pop-up windows containing information visualizations of the specified areas. On the analysis level, these regions-of-interest are connected to the simulation in real time. Local parameters can be modified through the associated pop-up visualizations which allows the user to immediately see the effects (Figure 1). Figure 1: Our probe-based visualization of an agent-based simulation, used to study the possible effects of development projects and other interventions in Afghanistan. Compared to the traditional agent-based simulation interface, the probe-based visualization offers multiple advantages. Without the ability to select localized regions, traditional visualizations for agent-based simulation depict information at a global scale. Similarly, the hundreds of controls for the parameters of the simulation are displayed without context. It is difficult to understand which slider controls what parameter, and at what geospatial locations. This example illustrates how a visualization can function both as presentation and analysis tools. Through high inter-