An architecture to support the co-exploration of data and history

Information visualization and analysis systems combine sophisticated algorithms with the power of human visual perception to rapidly find patterns in data. However this analysis is rarely applied to another complex dataset: the history of the analyst's discovery process itself. Visualizing and analyzing session history together with the data is a promising technique for understanding and modeling discovery, and an additional tool for analysts.; We propose a model for visual analytic systems to create a detailed, meaningful record of the evolving exploration session that users view and interact with as an integral part of their discovery process. Since we cannot anticipate all applications of our model, to highlight its power we provide an example of application to the problem of system evaluation, demonstrating some additional capabilities made possible by a dramatically different exploration support tool.; We have implemented such a system in a visual data exploration framework. Sessions from a group of analysts exploring a large medical database were recorded, including their data/history co-exploration. This effort provided insights about what history data to store and how to use it to support interactive exploration.; Many questions can be answered using such a system. When does a user return to previous states? What are recurring patterns? How does the type of data being explored affect the session? This research has not answered all of these, but it has elaborated a number of techniques and highlighted additional items needed for visual analysis to answer such questions.