Lecture on Sep 21, 2010. (Slides)


  • Required

    • Chapter 1: Information Visualization, In Readings in Information Visualization. Stuart Card, Jock Mackinlay, and Ben Shneiderman. (pdf)

  • Optional

    • Decision to launch the Challenger, In Visual Explanations. Edward Tufte. (pdf)

    • The Value of Visualization. Jarke van Wijk. IEEE Visualization 2005 (pdf)

    • Graphs in Statistical Analysis. F. J. Anscombe. The American Statistician, Vol. 27, No. 1 (Feb., 1973), pp. 17-21 (jstor)


jheer wrote:

Card, Mackinlay, and Shneiderman posit a dichotomy between "scientific visualization" and "information visualization" based primarily on the input data: physical "scientific" data (e.g., air flow over an airplane wing) often has a natural spatial mapping whereas "abstract" data (e.g., stock prices or an online social graph) require that spatial mappings be designed. However, my colleague Tamara Munzner aptly points out that "information visualization is not unscientific, nor is scientific visualization uninformative". To what extent do you agree or disagree with this distinction? Do you see it as helpful or hurtful to the study of visualization?

Card et al. also mention Larkin and Simon's study of people solving physics problems with and without the use of diagrams. If you are interested, I highly recommend reading their paper Why a Diagram is (Sometimes) Worth Ten Thousand Words.

yanzhudu wrote:

Many Eyes


It is a good place to experiment with different types of data visualization

barbagli wrote:

I was impressed by the Sphygmograph - I guess before it came out, all blood pressure measurements were invasive. Two interesting images in addition to the one from the slides:

http://www.phisick.com/a1msgm.htm (pictures of an actual device)

radialapp (similar to the picture from today's class but with an example of the curves drawn by the device)

andreaz wrote:

My take from Munzner's quote is that she wants to emphasize that Card, Mackinlay, and Schneiderman use a strict sense of the term "scientific," in that "scientific visualizations" are based on data derived from the natural sciences. I agree with Munzner's thought that information visualization is not necessarily unscientific; information visualizations can be created out of data from the social sciences or formal sciences, which are still sciences in that they are objective and systematic studies of an area of knowledge, but unlike the natural sciences, these sciences are not based on physical information with an underlying geometric structure. I believe that Card, Mackinlay, and Schneiderman differentiate between the two types of visualizations in order to emphasize the issues that arise when attempting to create visualizations using information without an obvious spatial mapping. I also agree with Munzner's second part of her quote, that scientific visualization is not necessarily uninformative, because scientific visualizations often show abstractions that have the potential to amplify cognition.

On an unrelated note, I believe the visualization about the sleeping patterns in babies is very elegant, but I don't agree with their explanation that infants are operating on a "natural human 25-hour cycle" until the 17th week. This was likely based on early research that suggested that people isolated from daylight and timekeeping preferred operating on a 25-hour cycle, but the results of this study were confounded by participants' exposure to indoor electric lights. More recent research has suggested that the actual human circadian clock runs on a daily cycle closer to 24 hours. The abstract: http://www.sciencemag.org/cgi/content/abstract/284/5423/2177

amirg wrote:

I agree that creating a dichotomy between "scientific visualization" and "information visualization" does help to highlight the issues that arise when there is not a natural spatial mapping, but I also think that following the natural spatial mapping can be problematic at times. For example, the Challenger diagram that depicts the different rockets side by side (in lecture slides and also Figure 1.6 in the reading) emphasizes the location of the O-ring damage within the rocket field joints. However, this representation obscures the actual relationship we care about when deciding whether or not to launch the rocket on a cold day, which is that between damage and temperature.

In general, while having a natural spatial mapping can be helpful because it may be more intuitive to visualize in some ways, it can also be limiting. For this reason, I find Card, Mackinlay, and Shneiderman's description of the ways that visualizations can be used to amplify cognition to be more useful (e.g. increasing the memory and processing resources available to users, reducing the search for information, and so on as shown in Table 1.3). Thinking about a visualization in terms of the cognitive function you want it to aid can help you understand the ways in which someone is likely to use the visualization, and can thereby guide both design and analysis. Which is not to say that this mode of thinking about visualizations doesn't have its own drawbacks…

rakasaka wrote:

In addition to crystallizing knowledge I think a term I find relevant is knowledge "packaging" in that in an age where time is considered such an important factor in grabbing and retaining someone's attention, efficiency is so critical - how can knowledge be packaged so that it can be interpreted in the shortest time possible.

And while Card, Mackinlay, and Shneiderman approach the idea of perception from a physical standpoint by dissecting the properties of an eye, I think it would be interesting to examine perception (and efficiency) from a global vs. local perspective - what may be easy to perceive for you may not be simply because I am Japanese, for example.

jdudley wrote:

The journal Nature Methods has a nice piece on design principles for data figures this week.

"Design of data figures" : Nature Methods : Nature Publishing Group


yanzhudu wrote:

The distinction between "scientific visualization" and "information visualization" seems artificial. The purpose of any visualization is to facilitate knowledge discovery and information processing. No matter the form (or technique) of the visualization method, the goal is to use our evolution-tested vision (and its associated processing capability) to its fullest. Visualization achieve its purpose when it enables human to process information better or discover new patterns. The rest is technical (nevertheless important) details in getting there.

With that in mind, the difference between "scientific visualization" and "information visualization" is really a question about what information and how that information is presented, which is a challenge for all type of visualization. There is no fundamental difference between these two.

gneokleo wrote:

Stuart Card, Jock Mackinlay, and Shneiderman make a good point when they say that the "right visual representation can make a problematic decision obvious". While the figure with the rockets has information that may not be visible in Tufte's visualization (e.g. location of the damage) this information is not necessary or vital to the problem that the scientists were trying to prove. Choosing the right representation can also reveal information that we may not have been aware of and i have certainly experienced this with data sets from various experiments. This interesting point also comes to show that sometimes convincing people about something is all about choosing the right representation. This would actually be a good experiment to see whether a person's opinion about something changes depending on how you represent data.

trcarden wrote:

After reading chpt1 information visualization i have a loose grasp on the different variables that go into a effective visualization. However the paper might benefit from a good dose of (interactive) data visualization itself. In addition to reading how different retinal stimuli effect our ability to understand and interpret graphs, why not also create a interactive visualization to demonstrate first hand the very principles outlined in the reading? They do include pictures of example interfaces but they just aren't as effective (imho) as being able to play with them as a method of learning about their design choices.

On the other hand the lecture did do a very good job explaining these topics again and in combination with some of Professor Heer's demos it filled in the gaps.

jbastien wrote:

I think the distinction between scientific information visualization is indeed very artificial. Case in point: Tufte's example of the O ring damage is nothing but scientific data visualization, yet his "optimal" representation is not at all a "classic" scientific visualization with a real-world mapping.

The distinction is artificially created because often "scientific visualization" means displaying directly observed/calculated information with a spacial mapping, whereas often "information visualization" will imply the visualization of second order effects or aggregations of data which often doesn't have a useful spacial mapping. Nothing precludes the display of second order or aggregate data for science.

What one must understand is that visualization is a tool used to answer a question, and often scientists try to visualize the effects their design has on its environment (e.g. "how does the air flow around this wing shape?"), and this is inherently physical.

The other issue I take with this is the assumption that real scientists build physical things. In a field dominated by computer scientists it's a fairly amusing generalization.

msavva wrote:

Card et al claim that "the purpose of visualization is insight, not pictures" where insight is defined as having goals of discovery, decision making and explanation. This definition seems to exclude the act of visualizing for aesthetic and artistic purposes. I think it is true that often the most successful and disseminated visualizations aiming to provide insight as defined above also end up simultaneously being exemplars of aesthetic beauty (perhaps there is even a correlation which would justify the pursuit of artistic perfection even when one is only interested in the functional aspects). The optional reading on the value of visualization briefly discusses this aspect but I also ran across a site which has interesting thoughts and examples on the subject of "Visualization and Art".

Following up from rakasaka's comment, I recently read an amazing article that purports to explain differences in the cognitive processing of images and text due to differences in the facial structure characteristics of Asian and Occidental people. It was an extremely interesting read albeit slightly far fetched at times. A link to the most pertinent section:


(I found the part explaining the differences in text flow direction to be extremely convincing)

selassid wrote:

In the critique to Tufte's argument, his scatterplot is deemed flawed because it is not displaying the physically relevant O-ring metrics, and this points out the larger need for context and understanding of the system being studied. Knowing the physical or systematic significance of variables is not always trivial and the fact that a convincing visualization can be created with misleading data especially highlights that extra care should be taken when using a visualization to come upon an interesting phenomena; if nonsensical transforms are taken of data, a nonsensical effect can be found. Knowledge about the topic being visualized is essential.

adh15 wrote:

I appreciated the distinction drawn by Card, Mackinlay, and Shneiderman between controlled and automatic processing (p. 25) because it serves as a guide for knowing when to augment graphics with textual labels. A good visualization should facilitate both kinds of processing thereby combining search and pattern detection with the opportunity for precise calculation. Topographical maps do this well by using color and contour lines to facilitate automatic processing, while labeling elevations, latitudes, and longitudes to facilitate more precise controlled calculations. I suspect that this powerful combination often yields visualizations that encourage a rapid and iterative cycle of question and answer among the people who view them.

jtamayo wrote:

One of the major purposes of information visualization is to communicate. This was clearly explained in the lecture, but only seldom mentioned in the reading. It seems to me that Card, Mackinlay and Shneiderman see information visualization as a dialogue between a human and a computer; after said human has found a satisfactory conclusion the process is essentially done, and all that remains is to "package the patterns found in some output product."

This view ignores how visualization tools can be used collaboratively to facilitate communication and shared cognition. It also ignores the difficulties of creating a visualization that is easy to understand for someone who is not deeply familiar with the data used to build it.

emrosenf wrote:

While reading the chapter, I kept thinking about vision in comparison to the other senses. One thing in particularly that struck me is that while I can remember the general outlines, trends, and low-resolution "themes" from the visual diagrams in the book and from class, I can't remember any higher-resolution specifics, like statistics.

When trying to memorize lists (like the cranial nerves), I always looked for mnemonics and found them helpful. Never did I try to draw a picture.

Is it possible that aural stimulus is easier to memorize at high resolution (that's why rhyming is popular), and visual stimulus is easier to memorize at low-resolution?

Are there techniques to emphasize specific statistics (absolute, rather than relative) in a visual diagram?

nikil wrote:

The standard structure of the data tables and the similarities to traditional relational data structure provide an interesting insight into the function of the data for each purpose. In the traditional view of the relational database the data is meant to be analyzed and manipulated. In the data table view, which flips the columns and rows, the data is used to restructure the table after examining the data and the patterns held within.

Perhaps instead of the traditional slice and dice methodology of OLAP http://en.wikipedia.org/wiki/Online_analytical_processing a technique like the restructuring of data visualization tables could provide better insight into the data.

Since this structure is required to be consistent for the real time use of the database on the web, we would have to do this offline or else figure out a way to have a middle logic layer to translate the data input query in real time to fit the new structure of the updated database to do the whole process while online.

Either way, (the second one being much more difficult to do), Analyzing the data in a different view of the schema which is not just simply sliced, diced, rolled up, or aggregated would be very insightful in visualizing new patterns in the data in the web source.

dlburke wrote:

Reading over the Herb Simon quote and thinking about how one of the goals of visualization is to trim down the information conveyed to that which is essential to answer whatever question is posed. The initial visualization of comparative brain power with its overabundance of labels scattered around or the Challenger rockets graph are great examples of how not to accomplish that. But a lack of information can also lead to disaster. And it seems that while the visualizer (?) might have his idea of what needs to be answered, it seems not unlikely that a reader may draw other conclusions, which were not intended to be pondered because of information eliminated to make the targeted questions easier. While it would be impossible to guess what every reader may be thinking, it seems that the designer should be concerned not only with the direct questions, but also with what additional conclusions the reader may infer in choosing to eliminate information.

jeffwear wrote:

For a book about developing expressive and effective visualizations, Card et al. commit several puzzling lapses of judgment in developing their own visualizations. Take, for example, Figure 1.15, showing knowledge crystallization. This diagram is extremely complex, with a number of difficult-to-distinguish arrows linking concepts. While these relationships might exist between stages, if the viewer cannot make effective use of them, then they ultimately detract from the diagram. In addition, each stage is annotated at length with sub-tasks. As the body of the accompanying section explains each stage in detail, these annotations could have probably been omitted for clarity's sake.

Furthermore, the diagram is mired in cartoonish kitsch - the stages appear to be represent the thoughts of some unnamed information-gatherer shown below the diagram. These thought bubbles are disorderly and non-uniform - several have elliptical bulges on their right sides. The cartoon person himself is distracting, as he has a giant misshapen arm with what looks like spikes on his hand. Surely his head could have been shown alone, if the designers insisted on pursuing the thought-metaphor - which itself may be a mistake. Portraying the stages as thoughts implies that the information-gatherer is or should be explicitly considering each stage as he goes about his task. This reframes the passage on knowledge crystallization not as a conceptual framework for describing people's activities but proscribing them, a stronger contention.

I would have liked Card et al. to have innovated when placing their visualizations within the text, as well. Visualizations are commonly distanced by a half a page from their description in the text, and in some unfortunate cases this gap is more than a page. At least reading one page at a time in digital form, it's difficult to keep both diagram and description in mind at the same time. Even if I could have two pages in front of me, it would still be preferable to have a smaller distance between related elements, to provide some link between the two (perhaps a faint line from "see Fig. x" to the visualization, or if this were in digital version, a hyperlink), or to provide a fuller annotation of the visualization so that the reader may linger there rather than have to return to the body text to find description.

anomikos wrote:

The Challenger case presents an interesting example of how the use of visualization enables people to come to better conclusions or just understand the underlying problem better. While reading the chapter and going though the overview of all the concepts that are part of the visualization theory I was thinking about other areas that this concept is applicable. I immediately though of newspaper (whether paper or online) articles where the author is using a very rich medium (natural language) to communicate ideas and opinions on a subject. Whereas the use of small diagrams is a common way to make the reading experience richer I couldn't help but wonder to what extend can the content of articles be substituted by visualizations. Especially in the digital edge where news come and go at an extremely fast pace being able to "read" articles by just viewing them would definitely be an interesting concept.

ericruth wrote:

I really like the point @rakasaka brought up about designing from a global vs. local perspective. It makes me wonder how encoding effectiveness may vary from culture to culture. For example, I wonder if Mackinlay's rankings, or other rankings, would hold up well for all humans from all backgrounds. Essentially, it makes me wonder what aspects of perception are inherently human, and what aspects might be influenced by our environment. I don't know if much research has been done in this area, but I could see this becoming more important as time goes on and cross-cultural collaboration increases.

hyatt4 wrote:

I found the quote by Norman (1993) (taken from Readings in Information Visualization) quite interesting. He says, The power of the unaided mind is highly overrated, and then goes on to say, By the invention of external aids: It is things that make us smart. The main authors, go on to discuss how visualization tools are one such useful external aid that helps us communicate our ideas as well as develop ideas themselves. I agree with this, but it makes me wonder how much someone's thinking would be effected by not having ever seen. I think it would make communication more difficult, but I do not think that it would hinder their smartness, unless the term smart by Norman is simply referring to ones ability to recognize external aids and communicate with others through those aids.

jayhp9 wrote:

@hyatt4 That is a very interesting point you brought up. It is a wonder as to how a blind person's thoughts might be shaped around the objects they perceive. I am inclined to say that the thought power of such people is superior to those who can see.

2 reasons:

1. They are not reliant on external aids to drive cognition. Whereas, people who can see often become so reliant on visual memory that we suffer withdrawal symptoms when the visual artifacts that usually aid us in thinking are not in front of us.

2. The thoughts of blind people often become much more sensory than those who can see. By this, I mean that blind people will try to experience things using their 4 other senses. Since they never actually get the perfect image of what something looks like in reality, they are constantly forming a picture, and redefining it whenever they find that their current understanding of a real object does not hold. In this manner, they are much more receptive to changes in their viewpoints, and may also tend to be more creative in their expression of ideas and emotions because they do not have to confine their understanding of an object to what it looks like.

Leave a comment