Handbook of Human-Computer Interaction Second completely revised edition M. Helander, TK. Landauer, P. Prabhu (eds.) 0 1997 Elsevier Science B. V. All rights reserved. Chapter 2 Information Visualization James D. Hollan Benjamin B. Bederson Computer Science Department University of New Mexico Albuquerque, New Mexico, USA Jonathan I. Helfman AT&T Research Murray Hill, New Jersey, USA 2.1 Introduction ....................................................... 33 2.2 Lenses As Information Filters .......................... 33 2.3 Pad++: A Zoomable Graphical Sketchpad ..... 34 2.4 Montage .............................................................. 37 2.5 Dotplot ................................................................ 39 2.6 Information Visualizer ...................................... 42 2.7 History-Enriched Digital Objects .................... 43 2.8 Toward A New View of Information ............... 44 2.9 Acknowledgments .............................................. 46 2.10 References ........................................................... 47 2.1 Introduction Computation provides the most plastic representational medium we have ever known. It can be -employed to mimic successful mechanisms of earlier media but it also enables novel techniques that were not previously possible. Computationally-based information presenta- tions promise to dramatically enrich our understand- ings as well as assist us in navigating and effectively exploiting rapidly growing and increasingly com@x information collections. In this chapter we survey a sample of recent information visualization research. Information visualization has a long history, dating to the earliest forms of symbolic representation, and can be approached from multiple perspectives, ranging across psychology, epistemology, graphic design, lin- guistics, and semiology to newer perspectives emerging from cognitive science. There are numerous introduc- tory surveys of information visualization. Examples include popular books by (Tufte, 1990) practical work on data visualization (Keller and Keller, 1993), and work applied to specific fields such as statistics (Cleveland and McGill, 1988). Information visualiza- tion research has grown dramatically in the last few years. Well-designed visualizations can be tremen- dously helpful but are still very challenging to create. There is an increasing amount of work on automating the production of visualizations (Mackinlay, 1996) and on providing tools to assist in designing effective inter- active information visualizations. The Sage and Visage work of Roth and his colleagues (Roth, Kolojejchick, Mattis, and Chuah, 1995) is particularly noteworthy. Our goal in this chapter is not a comprehensive survey of information visualization but rather to pro- vide a glimpse of current research and attempt to com- municate the exciting potential of new dynamic repre- sentations. To accomplish this we profile selected re- cent work from our research group and others. We then step back from the details of specific projects to dis- cuss what we see as the beginnings of a paradigm shift for 'thinking about information, one that starts to view information as being much more dynamic and reactive to the nature of our tasks, activities, and even relation- ships with others. 2.2 Lenses As Information Filters Consider the series of screen snapshots in Figures 1-4. They depict movable lenses (Stone, Fishkin, and Bier, 1994) that combine arbitrarily-shaped regions with fil- tering operators. They can be moved over objects to dynamically change the views presented. Figure I de- picts placement of a pair of lenses over a section of text. The upper lens highlights text with special prop- erties. In this example, it has highlighted Middle and Right and thus has provided additional information, regions that can be selected with the mouse, that would 3334 Chapter 2. Information Visualization Figure I. Pair of movable lenses. Upper lens reveals special properties of text and the lower lens shows tabs and formut- ting. Figure 3. Detail lenses. Here placed over the starting and ending location of a journey to assist in planning. normally be invisible. The lower lens similarly shows whitespace and formatting codes such as tabs (here in- dicated by triangles). The lens in Figure 2 allows the identification of fonts used in a portion of text. Figure 3 uses lenses to provide greater detail about sections of a map. In this particular example, lenses have been placed over the starting and ending locations of a journey to assist with planning. One, for example, typically needs much more detail about streets in the areas around the beginning and end of a journey. Figure 4 demonstrates a lens that one can click through to display the definition of a word. These examples of simple interactive information visualization techniques employ a metaphor based on physical lenses but allow for creation of lenses with dynamic functionality that goes beyond mere magnifi- cation. They are movable regions that provide alterna- tive representations of objects within an overall con- text. There are a number of research groups exploring the use of lens-like interfaces to support information visualization. The work described above comes from researchers at the Xerox Palo Alto Research Center (Bier, Stone, Pier, Buxton, and DeRose, 1993). Earlier work at the MCC Human Interface Laboratory investi- fo; each button: Figure 2. Font iaentification lens. Shows fonts used in re- gions it is postponed over. Selection The mouse has three buttons named LEFT, MIDDLE physical ds for each Figure 4. Definition lens. Clicking through this lens displays the definition of a word. gated similar interface mechanisms for selectively viewing information in large knowledge bases (Hollan, Rich, Hill, Wroblewski, Wilner, Wittenburg and Grudin, 1991). This technique, also termed lenses, provided for access to alternative perspectives on knowledge base entries. Related work on brushing to link data across different views has been very valuable for analyzing multi-dimensional data (See Cleveland and McGill, 1988). For the last several years we have also been exploring lens-like filters as part of the Pad++ dynamic multi-scale interface research effort that we describe next. 2.3