Reasoning in Spatio-Temporal Analysis - Theory, Provenance, and Applications

The aim of this dissertation is to increase our understanding of human reasoning by means of maps and other visualizations in spatio-temporal (S-T) analysis. The theoretical approach is that an improved understanding can be achieved by applying theories from cognitive science and psychology, by studying human reasoning in S-T analysis tasks, and by putting special focus on the role of spatial and temporal relations. This improved understanding will enable us to derive guidelines for the creation of representations, tools, and ways of working that better support human reasoning in various aspects of S-T analysis. Specifically, this dissertation is motivated by the need to document and communicate the reasoning processes behind analysis decisions, a need related to that of provenance in visual analytics. The contribution is threefold. First, this dissertation presents a human reasoning framework for S-T analysis. This framework emphasizes the role of bottom-up and top-down processes, the use of knowledge in the form of mental models, the qualitative nature of human reasoning, and the spatial, temporal, and thematic aspects of S-T data. Second, based on this framework and relating to provenance, four guidelines for the creation of visualizations that communicate the reasoning of S-T analysis are derived. These guidelines acknowledge that it is not enough to cover the outcomes of the reasoning, but that the whole process that led to these outcomes needs to be considered. Two visualizations created according to the guidelines are presented. Third, this dissertation also includes three more application-oriented studies. The two first are regarding map animation and consider how to improve its effectiveness as a tool for S-T analysis. The findings suggest that sonification can be beneficial for this purpose and that classification of events based on their temporal intensity facilitates reasoning and leads to more insightful animations. The third study regards domain knowledge acquisition in agent-based modeling, and presents a visualization approach to be used for this purpose that facilitates communication between modeling and domain experts. The framework provides a means for understanding human reasoning in S-T analysis and constitutes a solid base from which to start the development of new representations, tools, and ways of working. The guidelines provided represent a new approach, not only in S-T analysis but also to provenance research in general. The reliability of the findings is largely dependent on the reliability of theories of cognitive science, such as the mental model theory of reasoning. The role of intuitive reasoning in S-T analysis is a topic for further research.