|Content Overview of Information Visualization (both Editions) and Selected Tools|
|Review of Information Visualization (1st Edition)|
|Review of The Craft of Information Visualization (Bederson & Shneiderman)|
|Review of Elements of Graph Design (Stephen Kosslyn)|
|Visualization – New Controls and Applications|
By Gerd Waloszek, SAP AG, SAP User Experience – February 12, 2008
This review takes a personal look at the second edition of Robert Spence's textbook Information Visualization.
Information: Information presentation
Robert Spence is Emeritus Professor of Information Engineering at the Imperial College of Science, Technology and Medicine, London, where he has been involved in research into human-computer interaction for over 30 years. Much of his research has involved innovation in information visualization. (From book cover, adapted)
Seven years after the first edition of his textbook Information Visualization, Robert Spence published the second edition of his book, increasing the volume by more than 70 pages, although not the book's thickness – the new edition seems to be made of lighter and thinner paper. The publisher explains that the second edition has been updated as follows:
Nevertheless, a lot of what I have already written about the first edition of the book is still valid. Therefore, readers might also want to read my review of the first edition.
In his foreword to the second edition of Information Visualization, John Stasko from Georgia Institute of Technology, an expert in information visualization himself, remarks: "We are awash in data. ... How can we help people to make sense of all the data?" He answers his own question as follows: "By better organizing it, giving it a structure, and providing it with a meaningful representation, we are able to transform the data into a format that people can more easily and quickly explore. That process is what information visualization is about." Stasko adds that the field of information visualization has two aspects: first the supporting technical foundations for creating representations of data, such as math, computer graphics, user interface techniques, and second the human aspect, that is, an understanding of the perceptual and cognitive abilities of people.
Spence himself particularly emphasizes the second aspect, admitting that his view of information visualization may be somewhat different from the mainstream concept: "Unlike most advocates of information visualization I stress that its definition refers to the developing of a mental model of data." He continues: "Thus, visualization is solely a human cognitive activity and has nothing to do with computers. An insight or understanding is acquired." He highlights this view by pointing to the "ah ha!" reaction: The sight of a graphical encoding of data causes this reaction in the viewer, indicating that a useful discovery has been made.
According to Spence, this "activity need not involve a visual experience, as might be suggested by the term 'visual' in visualization, because sound and other sensory modalities – not only graphics – can be employed to represent data." Regrettably, Spence provides only a few examples of acoustic encoding from the health sector in his book (for another example, see Zhao, Shneiderman, Plaisant (2007): Listening to Chloropleth Maps: Interactive Sonification of Geo-referenced Data for Users with Vision Impairment; reference)
That said, the book of course mostly covers the first of the two aspects that Stasko mentioned – even though Spence has added new material dealing with the second, the human, one.
Like the first edition of Spence's book, the second edition is primarily aimed at undergraduate and masters students of computer science. In addition, the book is not only targeted at software engineers, but also at the "huge population of professionals ... whose activity can potentially be immensely enhanced through the application of information visualization." We shall see further down whether the second edition better fulfills the latter goal than the first one.
As already mentioned, the book begins with a foreword by John Stasko, followed by the author's own preface, in which he outlines the key issues of information visualization, comments on the book's target audience, and – because the book is a textbook – provides instructional advice.
The first two chapters of the book serve as an introduction: The first chapter defines information visualization and the second identifies the three principal issues underlying information visualization, namely representation, presentation and interaction. This is done in the form of a walk through using the Attribute Explorer, a visualization tool developed by Spence and his coworkers (see figure 1). This tool reappears throughout the book, acting as a leitmotif.
Perhaps it would be useful to recall Spence's definitions of the three above-mentioned key concepts:
I am not sure, whether all authors would agree with these definitions. Personally, I always have problems telling "representation" and "presentation" apart, even though I have often come across their definitions.
Figure 1: A Standalone Java Version of the Attribute Explorer (by Andy Smith, IBM; Screenshot by the Author); Note the Highlighted House that is Being Brushed to Linked Histograms
The following three chapters constitute the core of the book. Each of them is devoted to one of the three principal issues of information visualization:
Chapter 6 Case Studies presents five projects illustrating the issues involved in information visualization. Most of the examples present visualization techniques developed by Spence and his coworkers. As already indicated, Spence uses tools developed by him and his coworkers as illustrative examples throughout the book (see content overview and selected tools).
The book is complemented by a glossary of terms from the field of information visualization, an appendix referring to the videos on the accompanying DVD, extensive references, and an index.
For a more detailed overview of the contents of the book and a comparison with the first edition, see the content overview and selected tools.
I assume that a few words about selected presentation techniques will help readers better understand the concepts and techniques presented in the book. From the book, I learned that "overview and detail" is not the same as "focus and context." While the first approach shows both the overview and selected detail information side-by-side, the latter approach includes the focal information and its context in one coherent presentation: The focus is shown with more detail, whereas the context exhibits decreasing detail the farther it is from the focus area. Distortion, suppression, or a combination of both, can be used for achieving this effect. To envision distortion, think of a magnifying glass with smooth borders that can be moved across a larger image. Other approaches to using distortion, such as the hyperbolic tree, are based on complex mathematical calculations for distorting the viewing area. Spence defines suppression as the "removal from view of a representation of data, according to some automatic or manually chosen criterion." Suppression serves a similar goal as distortion, namely to balance local detail with global context. As an example, in an exploded view of an engine parts that are irrelevant for repairing a certain fault can be hidden from view. Finally, the zoom and pan approach provides the smooth continuous movement (pan) of an enlarging viewing frame (zoom) over a two-dimensional image of greater size. This sounds similar to what I have just said about distortion. However, the transition from the enlarged area to the global view is abrupt, and the viewing frame hides parts of the larger area. Thus, this approach more closely resembles the movement of a real magnifying glass over, for example, a map.
For fighting time limitations, Spence and his coworkers developed the technique of rapid serial visual presentation (RSVP): Images are presented in rapid succession (typically, each image is shown for 100 milliseconds) instead of concurrently; users have to identify items from the rapidly presented set. This techniques trades space for time, utilizing the human ability to identify images in large image sets with high success rates.
According to Spence's definition, interaction serves mainly to change the view of presented data. For continuous information spaces, that is, for continuously varying data, a continuous interaction is most appropriate. For example, this type of interaction allows users to change certain values in an electronic circuit and observe the effects of their manipulations. For discrete information spaces, which is a common case in everyday life, a stepped interaction is more appropriate. As an example, hypertext or the Web is a discrete information space; here, navigation is performed by clicking links – a stepped interaction.
In the context of stepped interaction, Spence discusses three interesting concepts: sensitivity, residue, and scent. Sensitivity denotes a single movement in discrete information space and the interaction required to achieve it. Formally, it can be described as a 2-tupel (SM, SI), where SM denotes a single movement and SI the interaction needed to achieve it. This formal notation may look more complex than the concept really is, as the following example shows: Links (or buttons) indicate that a jump to new page is possible and that clicking them achieves the jump. Thus, links answer the two basic questions: "Where can I go?" and "How can I get there?"
Sensitivity information points to only one possible step in an information space. Residues go beyond that: They provide users with an indication of distant content in the SM encoding, that is, they point to content that is more than one movement away, allowing users to look ahead farther. Scent denotes "the perceived benefit associated with a movement in information space, evaluated following interpretation of one or more cues." In other words, scent provides users with a notion of whether it is worthwhile to pursue a path in information space. Recently, this concept has been popularized by Jared Spool in the context of Websites (the "scent of information").
Finally, I would like to refer to the technique of brushing. According to Spence, brushing is "a change in the encoding of one or more items essentially immediately following, and in response to, an interaction with another item." But that explanation may be somewhat difficult to understand. Therefore, I will provide an example from Spence's book: Think of objects that can be classified in different ways, such as houses that can be classified according to price, number of rooms, and time to drive to work. The house data can be displayed in a 3D-scatterplot, a 2D-projection of it, or three 2D-projection planes, which show the relations for two of the three dimensions. Users may pick items in one scatterplot and highlight them. Brushing means that the same items in the other two plots are also highlighted immediately. Spence describes this as: "We are brushing houses from one plane to the other two." All in all, brushing allows users to explore the effect of changes in one parameter on the relation between the other (here, two) parameters (see figure 2). Among others, Spence's Attribute Explorer uses this powerful technique (see figure 1 above).
Figure 2 : Brushing Two Houses (Marked in Red) from One Plot to the Other Two. (After Spence, 2007)
At first glance, it may appear that the second edition is just a scrambled version of the first one. However, while most of the old content has been retained and rearranged, a lot of new content, as I have already mentioned, has been added and other content updated.
This leads to the question of how effective the new structure of the book is. To me, the new structure is much more logical and clearer than the old one, because the book has been organized into general categories. Nevertheless, I am not completely convinced of the new structure. General terms do not always make things clearer, particularly, if people have problems understanding the difference between terms, such as representation and presentation, as I have from time to time. To refresh my impression of both editions of the book, I quickly scanned through both books page by page. And I got the feeling that the first edition looked more interesting than the second one (and maybe a bit more chaotic...). How could that be? Doesn't that go against all logic and reason? My gut feeling is that the revised structure has made the book duller, too much of the same thing is grouped together. For example, reading five case studies in a row can be more exhausting than reading them interspersed with other text. Perhaps the book would be more interesting to read if the case studies were placed at the end of each chapter, instead of being consolidated in a separate chapter.
When reviewing the first edition, I wrote that it was difficult to find certain topics in the table of contents. Readers had to either consult the index or scan the book. This situation has not changed much, and the new categories representation, presentation, and interaction may not be an improvement for practitioners. Spence has designed his book as a textbook, and this shows. In fact, adding exercises reinforced the textbook character of the book. Readers who are looking for a cookbook for information visualization may still have to look for other information sources.
As already mentioned, Spence has added new exercises for students to each chapter of his book. He suggests that teachers use them in group exercises as the final activity of course modules. I must admit that I did not try to solve the exercises because I am usually very weak at this.
The book is also accompanied by a DVD with 37 movies in Apple QuickTime format (on Windows computers, a QuickTime installation may be required; some videos are also provided in other formats). The videos are well documented in the appendix of the book. The file formats and durations are provided, and it is noted whenever a video lacks sound. In some cases, the date of the video is also provided; if not, you may be able to conclude it from the references or comments. A handy feature also deserves mention: In the main text, symbols on the page margins indicate that a relevant video is available (the video number is given). The videos can be accessed from an HTML page (videos.htm) on the DVD. Regrettably, there is a small error on the page. When you try to access video 5, you load video 35. However, you can find video 5 in the respective folder on the DVD (video 35 "dust magnet" appears as video 5, too, but the correct video 5 is also on the DVD).
By the way, when I first tried to access the book's companion Website, I landed (or was stranded...) on a commercial information portal of unknown origin. Having recovered from my bafflement, I noticed that I had written "person" instead of "pearson" in the URL. However, on the correct site, I was not able to find any materials for students. Access to instruction materials was available but only for instructors at schools that are accepted by the Pearson Website (registration required).
I have already discussed the new structure of this edition and my personal feelings about it, and I also touched on the usefulness of the book for practitioners. In my opinion, the book could be improved in this respect in a further edition. It might provide more guidance for practitioners, be it in the form of selection tables, check lists, or something else to find appropriate visualizations for typical applications.
The book boasts a multitude of colorful illustrations, many of which have been updated in the second edition. As many visualization techniques involve dynamic aspects, 37 videos are provided on DVD to aid the understanding of the dynamic approaches. However, there is still one piece missing. Maybe it's because I am left-handed, but I never profited much from people demonstrating procedures to me – I need a hands-on approach. Psychology also suggests that we profit most from doing things ourselves (the so-called effect of "doing"). Particularly, for the Attribute Explorer, which has been developed by Spence and his coworkers and is used throughout the book for illustration, I would have appreciated a version of this tool on the DVD, written in Java for platform-independence and running in a Web browser. Using Google, however, I found an applet and a stand-alone Java version on the Web developed by Andy Smith and colleagues from IBM UK who had worked with Spence (see references).
Finally, since Robert Spence points out that information visualization is not restricted to the visual mode, more examples demonstrating the use of the acoustic or haptic modes would be welcome and instructive. Such an addition would also support the universal usability case.
As with the first edition, I highly recommend this book to students and professionals working in the user interface domain who are interested in information visualization and the role of human information processing. However, I am still afraid that the author may miss parts of his intended audience, particularly managers and practitioners who need fast and easy advice. The author told me that he is working on a third edition of the book. So, there is still a chance to approach and win these groups as readers – the book and its topic really deserve it.
See also on the SAP Design Guild
Note: Both pages are "bonus" additions by the review author.