Omnigram Explorer: A Simple Interactive Tool for the Initial Exploration of Complex Systems Tim Taylor, Alan Dorin and Kevin Korb Faculty of Information Technology, Monash University, VIC 3800, Australia tim@tim-taylor.com alan.dorin@monash.edu kbkorb@gmail.com Abstract We describe the design of Omnigram Explorer (OMG), an open-source tool for the interactive exploration of relation- ships between variables in a complex system. OMG is de- signed to help researchers gain a holistic, qualitative under- standing of the relationships between variables in their data at a preliminary stage of analysis; such exploration might high- light interactions that warrant further quantitative investiga- tion using other tools. We illustrate OMG’s use on real-world data, and also describe its potential as a tool for communica- tion to non-specialists. Introduction When working with models of complex systems, visualisa- tion tools can be invaluable in helping researchers gain an understanding of the system’s behaviour. Visualisation can be used at all stages of the research process, from the earliest stages of design through to the eventual communication of results to a variety of different audiences (Dorin and Geard, 2014; Grimm, 2002). The work reported in this paper was developed in the con- text of a collaboration between the authors and a group of epidemiologists at the University of Melbourne. 1 The epi- demiologists use a variety of different modelling techniques in their work, ranging from agent-based models and other simulations to more traditional mathematical approaches. Their models often involve several dozen independent and dependent variables. We wished to develop an interactive tool that would al- low our colleagues to gain a quick qualitative understanding of their models at the initial stages of analysis, and high- light features to be investigated in more detail. The resulting tool, named Omnigram Explorer (which we abbreviate as OMG), is described in this paper. In the following sections we review relevant previous work, describe the principles that drove OMG’s design, describe the tool’s main features, and give examples of using OMG on real-world data. 2 1 See Acknowledgements section for details. 2 Various supplementary materials are indicated in the pa- per. These are available at http://www.tim-taylor.com/ omnigram/ecal2015/. Omnigram Explorer is a free, open-source tool devel- oped in Processing. 3 The source code, binary executables (for Windows, Mac and Linux), documentation and related materials are available at http://www.tim-taylor. com/omnigram. Previous work Traditional approaches to visualising the relationships among multiple variables include the Scatter Plot Matrix (SPLOM) and Parallel Coordinates (Heer et al., 2010). These are widely used, although neither is without problems. For example, SPLOM visualisations focus on pairwise rela- tionships between variables (Kosara et al., 2003), and the effectiveness of Parallel Coordinates can greatly depend on factors such as the linear order in which data dimensions are plotted (Zhang et al., 2012). In addition to representation, interaction is an increas- ingly important aspect of information visualisation systems (Ward and Yang, 2004; Yi et al., 2007). For the kinds of visualisation problems we wished to address in our project, relevant and interesting early work was produced by Spence and Tweedie (1998). They developed a tool called Attribute Explorer in which multivariate data was presented as a set of histograms, one for each variable. Each histogram displayed the distribution of values in the data-set for its associated variable. The user could select subsets of data by adjust- ing a slider under a histogram. The subset of data points so selected was represented in the other histograms using a specific highlight colour (a technique known as linking and brushing). The tool described in the current paper, Omnigram Ex- plorer, took inspiration from Spence and Tweedie’s work as a starting point. We added a variety of novel and principled extensions (as described in the following sections), and have made it available as a free, open-source tool. Design principles An effective data visualisation should provide an intuitive way for the user to gain insight into the organisation of that 3 http://processing.org Tim Taylor, Alan Dorin, Kevin Korb (2015) Omnigram Explorer: A Simple Interactive Tool for the Initial Exploration of Complex Systems. Proceedings of the European Conference on Artificial Life 2015, pp. 381-388 DOI: http://dx.doi.org/10.7551/978-0-262-33027-5-ch068