14 Epilogue ................ 207

The postscript version of this chapter.

Table of Contents.

Chapter 14



One of the key challenges facing the emerging field of information visualization is to provide users with a variety of visual abstractions and powerful ways of linking these representations. It also needs to enable users to visualize how they have interacted with the information. Each of these abstractions will have a semantics associated with it, because it enables users to create meaning or gain understanding that has certain defining features in terms how users interact with the data. The way users choose to visualize the information and the actions they perform are valuable sources of information that also need to be understood by the users. Hence, we need to find ways to abstract and visualize this meta-information. We envision a future where users will be able to create and interact with a hierarchy of visual abstractions, where a higher level reflects and summarizes the actions and semantics of the lower level. This vision is still very vague, but it could provide us with a "golden carrot" that will eventually lead us to productive discoveries.

The work presented in this thesis grew out of such a vague vision: we wanted to develop a visualization that would give us a compact view of how the contents of a library are related to our specified interests. We wanted to be able to juggle as many different interests as possible at the same and use visual grouping principles to display all their relationships in a compact way. We wanted a representation that allowed us to focus on specific relationships without forcing us to abandon our sense of overview. We wanted to develop a representation that would provide us with a global, yet locally detailed overview of the contents of the database.

What has been the outcome of this vague vision? What did we stumble over as we were busy chasing after this vision dangling in front of our mind's eye? We have developed a powerful visual representation that uses a simple visual analogy to enable users to deal with some of the complexities involved in information retrieval. We have presented a novel representation that not only can be used to visualize arbitrarily complex Boolean queries or weighted queries, but also the vector-space retrieval approach and its related partial matching methods. We have developed a visual framework that can accommodate the major retrieval methods. Further, we have outlined how the InfoCrystal allows users to move effortlessly between different ways of accessing and viewing information.
We understand the InfoCrystal as a possible piece in the mosaic of emerging information visualization tools. The purpose of this thesis has been to demonstrate these capabilities in the form of a prototype and hopefully to have presented it in such a way to inspire others to build on the presented work.

We have stated that the way we process visual information has been an inspiration for creating the InfoCrystal. On the one hand, it only makes sense to turn to our understanding and evolving theories of the human visual system to give us insights into how to create effective visualization. It is precisely the human visual system that we are trying to engage in the process of gaining an understanding of large information spaces. This is the simple and straightforward motivation. On the other hand, there is a deeper reason to try to learn from the way we process visual information. The human visual system uses a multitude of representations to arrive at the perceptions of the physical world that we take for granted. There are bottom-up as well as top-down processes involved in this sense-making process, where these processes exploit the regularities of the physical world [Marr 1982]. It is our intention to suggest that there could be a parallel between the way the human visual system processes information and the type of information visualization environment we need to strive to create. Like our visual system, we need to provide users with a multitude of diverse representations and visual abstractions and with ways to link them that are powerful. The InfoCrystal represents one of these needed abstractions and it could be used to moderate the communication between different representations. To use an analogy, we want these links to be able to speak a language that goes beyond revealing correspondence (i.e., we want to move beyond highlighting corresponding points in the linked representations), but that is able to issue more complex commands that have the effect of transforming the way the linked representations are visualized.

There is a growing trend to hide from users the complexity of the methods used to accomplish a task. We have to strive to create visualization systems that find a balance between the ease with which the information can be consumed by our visual system and how to bring analytical and reflective cognitive processes to bear on the sense making process. Information visualization is meant to stimulate and guide our sense-making processes, without however to completely succumb to our established ways of perceiving the world. We are not in a habit to question what we see. On a deeper level, we rarely question what we believe. We are entering an age where our ability to make sense of large and diverse information spaces will become a key competitive advantage. The whole notion of database mining is to find the "eternal needle in the haystack". If we only knew what to look for or where to look ! Our expectations and beliefs play such an important role in how we perceive the world. How do we have to frame the information to be able to perceive potential valuable patterns ? The future seems to require from us to be able to shift our point of view at an increasingly rapid rate. We need to be able to change the context within which we explore information not just occasionally but on a constant basis. The constancy is change. The resulting challenges for information visualization are many-fold to support complex data explorations. First, we need to create visual abstractions that are easy for users to learn how to use. Second, we need to provide users with tools that enable them to create new visual abstractions that capture what they have learned. Third, we need to create visualizations that are able to learn from how users interact with them. We envision a self-organizing visualization environment, where users are integrated in the loop both as a source for learning and as participants that direct the learning process.

Information visualization faces the challenge of having to invent visual abstractions that capture the complexity of the information spaces that users are increasingly required to understand, using design principles that lead to simple and consistent tools. In short, one of the key tasks of information visualization is to increase the simplicity with which users can deal with increased complexity. This challenge is a reflection of the exciting opportunities for invention that the field of information visualization offers.
I will be back.