The Purpose of Visualization

From CS 294-10 Visualization Sp10

Jump to: navigation, search

Lecture on Jan 20, 2010




  • Chapter 1: Information Visualization, In Readings in Information Visualization. Card, et al. (pdf)
this PDF is quite janky; it's got some weird duplication thing going on in the middle of each page, and the bottoms are missing. (Pauloppenheim 09:43, 23 January 2010 (UTC))

Optional Readings

Maneesh Agrawala - Jan 01, 2010 11:30:22 pm

What do you think of Card et al.'s distinction between "scientific visualization" that deals with physical scientific data (e.g., air flow over an airplane wing) and often has a natural spatial mapping versus "information visualization" that deals with abstract data (e.g., stock prices or an online social graph) and usually requires designers to assign the spatial mapping? To what extent do you agree or disagree with this distinction?

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 recommend reading their paper Why a Diagram is (Sometimes) Worth Ten Thousand Words.

Jon Barron - Jan 23, 2010 06:07:50 pm

I found the most interesting contribution of this paper to be the "Cost-of-Knowledge Characteristic Function" on page 16. It's both formal and well-defined, and actually seems like it would be very helpful in analyzing visualizations. One complaint with this metric is that it does not differentiate between precision and recall, where precision is "how fast can I find the exact piece of information I'm looking for", and recall is "how fast can I be generally informed about the data".

Chetan Nandakumar

Card - I like how visualization is described as able to amplify cognition. The details of how this happens are well mapped out in Table 1.3. The two specific elements that jumped out at me was that visualization can leverage the power of perceptual system to process information and not solely rely on the cognitive system. In addition, by more effectively organizing information, it reduces the load on working memory.

As for the author's distinction between scientific visualization and information visualization, I certainly think a distinction needs to be made but I don't know if it's as fundamental as the author argues. The main distinction is that when a visualization is not about a physical process there are more degrees of freedom in how the data can be visualized and also more purposes as to why one is interested in the data. With a physical process the purpose is to understand how the physical system works and therefore the visualization is inherently more directed.

Pauloppenheim - Jan 24, 2010 06:03:20 pm

  • Q: "What do you think of Card et al.'s distinction between "scientific visualization" that deals with physical scientific data (e.g., air flow over an airplane wing) and often has a natural spatial mapping versus "information visualization" that deals with abstract data" A: most of that section sounds like bullshit.
    • quote: "Scientific visualization is visualization applied to scientific data, and information visualization is visualization applied to abstract data."
    • Wait, lemme catch up with you. What is the difference in the definition of "information" and "data"? I would say those words are synonyms, but by the logic of the authors, Scientific visualization : scientific data, and information visualization : abstract data. So information = abstract? I would imagine the implication from the wording of the science one would be information visualization : information data, which makes no sense. Hence, their distinction is word salad.
    • aside from the snarky "hey guys, write a little better" commentary, I do agree in some sense that the inherent geometric relations of physical systems give a bit of a natural nudge to those visualizations, insofar as there is a clear geometric property of the physical system. I can't imagine quantum mechanics or the 6-dimensional dances of bees are much easier to depict in a 2D projection than any other complex data.
  • Now that I'm done, I think the general thrust of "we use visualizations as a hack to make us smarter, and must be careful to avoid becoming stupider" is awesome. I think the approach (focusing solely on abstract data vis after discussing "scientific vis" as something different) is somewhat stilted. (outdated?) The approach discounts visualization methods of abstract concepts that closely match physical systems, which I find to be the most useful. For instance, the operating systems diagram for an LRU cache is typically a circle that marks slots. That diagram method isn't discussed in this survey, which mostly covers typical 2d graphs after asides about more unconventional methods.

All reading notes here, beware my grump.

Mason Smith - Jan 24, 2010 08:52:58 pm

I'm not convinced that the distinction between information and scientific visualization is appropriate, or, at the least, appropriately named. Lots of 'scientific' data (solubility of a liquid as a function of temperature, stress-strain curves, etc.) represent concrete measurements with no obvious spatial component. I think spatial visualization and abstract visualization might be better terms, but even that exaggerates the distinction. Even a dataset with a natural spatial component might still be ambiguous in terms of visualization; the discussion of various globe projections is a good example of this. Similarly, a purely spatial view of a spatial dataset might be unfavorable. For instance, a simple 2D contour map or 3D relief map of the tide phases in Figure 1.9 would be more faithful representations, but those methods may not reveal the anphidromic points so readily.

I think Card is implying that scientific / spatial data is somehow easier or more natural than abstract data, but as I have tried to demonstrate above, I don't think it's so straightforward.

Jiamin Bai - Jan 24, 2010 11:28:04 pm

What do you think of Card et al.'s distinction between "scientific visualization" that deals with physical scientific data (e.g., air flow over an airplane wing) and often has a natural spatial mapping versus "information visualization" that deals with abstract data (e.g., stock prices or an online social graph) and usually requires designers to assign the spatial mapping? To what extent do you agree or disagree with this distinction?

I feel that the distinction that Card et al makes is fair. There are definitely physical scientific data sets that are impossible to map or visualize in their natural spatial mapping simply because of their high dimensionality (light fields and light transport for example). However, more often than not, there is likely a simple way to visualize them (in the way it was collected) that allows humans to visually map the data presented to the actual physical experiment.

By grouping the "information visualization" as another group, what he is suggesting is that it leaves room for designers to optimize the way the data is presented. How this is done will severely affect the effectiveness of the presentation.

Therefore I would say that the distinction would only make sense for low dimensional data collected from the physical world. For high dimensional data from physical experiments, it requires ingenuity from designers to make it an effective presentation which would make it not much different from abstract data.

Mila Schultz - Jan 25, 2010 11:31:16 am

Card's argument for the distinction between scientific visualization and information visualization, while well-made, failed to convince me that the two are obviously separate. Card's terminology is questionable; scientific data does not always have an obvious spatial mapping, and data from non-scientific contexts can have a physical spatial component. Additionally, the spatial data could benefit from being displayed in a way more aligned with his definition of information visualization. If the field is truly divided as he suggests, it seems that scientific visualization could unnecessarily be tied to the original spatial mapping, and information visualization could be come unnecessarily abstract.

I particularly enjoyed Card's section on knowledge crystallization and external cognition. He clearly showed how visualizations can enhance memory and "think" for the viewer, and I have experienced this while studying physics.

Akshay Kannan - Jan 26, 2010 08:40:31 am

I agree in the distinction between "scientific visualization" and "information visualization," especially since scientific visualization is realistically modeled. While there are still a variety of means by which this physical model can be represented (degree of detail, points of closeup), the actual visualization resembles its appearance in real life, and therefore the various representations of the object/concept are finite.

On the other hand, the manner in which abstract information is visualized can completely change the purpose and meaning of the visualization, as there are infinite possibilities by which a given set of abstract data can be visually represented. While a good visualization can make patterns apparent and allow the reader to draw solid conclusions, a bad visualization can choose an unfit representation for a given set of data and consequently provide no useful information. I found it especially interesting how properties of human perception can be exploited in visualization techniques to enhance their impact.

Ebby Amirebrahimi - Jan 27, 2010 12:14:56 am

I certainly understand why Card would make the distinction between information visualization and scientific visualization. Information visualization attempts to associate some sort of picture to an abstract, intangible set of data. Scientific visualization, on the other hand, attempts to simplify the depiction of something that has a physical form in nature. So they certainly could appear to create a dichotomy. Nevertheless, I don't believe this distinction is fully founded. While they are describing different things, the goal for both types of visualizations is to aid understanding, and that doesn't require an exact physical, spacial depiction of something scientific. Describing the quantum behavior of matter is certainly not aided by a true-to-nature depiction, but requires something more abstract and creative. Similarly, visualizing abstract information may benefit from comparison to physical concepts.

Kerstin Keller - Jan 27, 2010 01:13:33 am

I also agree that we can distinct beetween "scientific visualization" and "information visualisation", and while they might be distinguished by the data and the appearance of the visualisation, both types of visualisation serve the same purpose: to help the user understand what the data tells, no matter if it is scientific or not.

Apart from this, I don't quite understand what is meant by "A formal treatment has the virtue that it is prescise, which is critical when discussing data, because subtle differences in data often result in large differences in visualization choices." (p. 17)

Does Card talk about differences in data value or data structure? I don't see why little change in data result in large differences on the visual side, since there might be multiple possibilities for a good visual structure.

Stephen Chu - Jan 27, 2010 02:13:53 am

I found the multiplication example, though simple, both insightful and entertaining. Maybe we were just lazy on Monday, but I thought it was funny how many people, including myself, tried to crunch the numbers in our head. Besides being inefficient, it's kind of embarrassing that we avoided using visual techniques in this class. I liked how this reading clearly stated why visualizations can enhance cognition: avoiding search by grouping related information, increasing memory, using perceptual inferences easy to understand, etc. However, if visualizations are so useful to us, why do we sometimes ignore them when solving problems like double digit multiplication? From this reading, their helpfulness is obvious, but not always recognized. How do we become more visual learners/problem solvers? What influences us to avoid or use visualizations?

Sara Alspaugh - Jan 27, 2010 08:22:15 am

Regarding Card's distinction between scientific and abstract data, I agree with Mason Smith and others that (a) this distinction would be perhaps better described as the difference between spatial and non-spatial data and that (b) even then, it is not always a clear-cut distinction, and many examples can and have been given for this (one of my own being physics data from which scientists try to create maps of the universe or depictions of an atom -- these clearly have a spatial component and are scientific but for these and other kinds of data on the very large and very small scales with which humans have no direct experience or intuition for, representations of them are just as abstract as any non-spatial economic or political data would be, if not more). Regarding Card's argument that visualization aids cognition, this is true, but it is not the case that without visualization cognition is made so difficult as to be impossible. The particular example I am thinking of makes me glad that he pointed out that other physical representations can also provide abstractions that also aid cognition, for instance, in the tactile realm. The particular example I am thinking of is the blind mathematician who created a visualization of the very difficult to visualize process of sphere eversion. There is a really interesting article about him and other blind mathematicians and what they think the differences between the way they think and they way sighted mathematicians think are and it is located here: < >.

Timothy Wheeler - Jan 27, 2010 10:12:16 am

I disagree with the scientific vs. information terminology used by Card et al. Scientific data is obviously included in the general notion of information, especially if we consider the idea that all measurements are inexact and our records contain other information, such as the accuracy of the instruments used to collect the data and the skill of their user. However, I think that special care should be taken to preserve the inherent properties of spatial data. For example, if unequal horizontal and vertical scales are used to plot spatial data, the resulting visualization may distort or misrepresent the relationship between the data. Some visualizations, such as streamlines indicating flow over an airplane wing, rely on the viewer's visual sense of proportion and scale to be successful. I think the real issue is graphical integrity in spatial visualizations.

Prahalika Reddy - Jan 27, 2010 01:59:18 am

I believe there is a distinction between what Card calls "scientific visualization" and "information visualization" in terms of the nature of the data that is used but I don't think the names or definitions are as clear or accurate as they could be. The term "Information visualization" is so broad that it doesn't clearly describe the concept of "abstract data with no obvious spatial mapping". In addition, it's not clearly explained what an obvious spatial mapping really is and which forms of data do or don't have such a mapping. For both forms of visualization, the data is converted to a visual in the end, whether or not the data is concrete or abstract or there is a clear mapping or not. The end product may be inherently different, the type of visualization may change according to the data, but the process of creating the visual should be similar so I'm not sure why there needs to be such a specific distinction of what type of visualization is occurring.

The rest of the article was interesting to read. The concept of information chromatography was new to me. I haven't seen that before, and I would like to see more examples of it. In addition, knowledge crystallization definitely seems useful for decision-making.

Shimul - Jan 27, 2010 12:37:39 pm

The notion of "Abstract" in the article is, in a paradoxical way, abstract. The term has been used with a little vagueness and thus the argument suggesting the distinction between "information visualization" and "scientific visualization" is not convincing enough. I think the author has tried to distinguish between the "easily perceivable" versus "not-so-easily perceivable" data visualization instead. Some kinds of data need deeper analysis than others, making them less cognitive and more abstract. This may not necessarily be stock prices, but also the 10-dimensional spaces studied in mathematics, which falls under "scientific" data. As I was reading the article, especially the topic on "perception", I realized how limited our resources are due to the limited capabilities of our senses, like the eyes. If we could invent visual aids that could give us superhuman vision for instance, we could expand the field of visualization by the same intensity.

The treemap discussion in class (and the treemap in Figure 1.37) were especially interesting visualizations.

Priyanka Reddy - Jan 27, 2010 11:58:16 pm

I think the point that Card et al. is trying to make by differentiating scientific visualizations and information visualizations is a good one. However, I think the labels are not the most accurate. Although a lot of scientific data has a natural spatial mapping to it, a good portion of it, such as temperature, weight, etc does not. Whatever the labels might be, understanding what information you're trying to convey with the visualization is the most important. Physical representations of physical data make sense if you're trying to show how the physical system works (ie. air flow over an airplane wing), as Chetan Nandakumar also mentioned, but abstract representations of this physical data might make more sense if you're trying to compare across systems (ie. wing drag with different wing shapes).

I also like the example of visual aids in performing multiplication. What was really interesting to me was the idea that writing the intermediate results in neat columns minimizes visual search. Usually when I think of using pencil and paper rather than doing it in your head, I assume the paper is helping mostly to remember the intermediate results, in which case we could have written all the intermediate results in one long list. Instead, the location of the numbers gives us a way to quickly understand what the number means. It's interesting to understand why this simple method we've used for so many years feels so intuitive.

Danielle Christianson - Jan 28, 2010 07:48:01 am

I think the distinction between scientific visualization and information visualization is valid in a general sense, i.e., it is useful for thinking about levels of abstraction. I find myself, in thinking about financial or business information, mapping the action of a financial transaction or business decision to a person making the action. Maybe the physical link in information visualization exists but it is most often not important or relevant to the main concept whereas in physical visualization it often is at least somewhat relevant.

I found interesting the discussion relating the human eye structure / function to the different types of visualization (spatial processing vs. object properties and controlled vs. automatic processing). The authors imply that these aspects can be exploited to reduce the cost-of-knowledge characteristic function by parallel processing of the information by the brain -- neat!

I appreciated the breakdown of visualization into discrete steps (raw data, data table, visual structure, views). I typically skip the explicit data table step and think it might be useful to consider in optimizing a visualization.

One concern I had was the authors definition of visualization that included "computer-supported". While I agree that computers are great and offer a lot of opportunity in visualization, I don't think they are necessarily required. Also, I wish the authors had presented their final table summarizing the chapter at the beginning -- the organizational aspect of the chart helped me to synthesize the many concepts of the chapter. It would have been nice to have that organizational structure to place concepts when encountering them.

Boaz Avital - Jan 28, 2010 04:16:18 pm

I actually rather like the distinction between what the author calls scientific visualization and information visualization. The actual name "scientific" visualization may be more of a historical artifact of the visualization of physically-based data than anything, and it may not apply nowadays when a perfectly valid physically-based visualization is "Best places to get coffee", something few people would consider scientific. Aside from that, I think it can be incredibly important to distinguish visualizations of data based in the physical world and visualizations of abstract information. It's possible that people have expectations when viewing physical data, they may expect the visualization to map in an obvious way to the physical source from which it was construed. It may be easier to understand and analyze it that way, which is the goal of any good visualization. Meanwhile, visualizations based on abstract data has different considerations. It could be that people are used to viewing certain data in certain ways (eg stock data in line graphs). However this field could be more cognitively flexible for people in terms of what ind of visualization they can easily understand.

I also find the broader idea of "perceptualization" interesting. Adding extra dimensions to the way we view data is tricky but has the potential to be very useful if the use is correct and well thought out.

Jonyen - Feb 07, 2010 10:08:46 pm

Reading through this article, it's nice to know that there has been quite a bit done in the realm of information visualization, but I can't help but wonder what will be the future in this field. More important yet, what about the issue of information manipulation? I think that is an important area to consider as well because we must do more than just intake data and understand it. We need to be able to apply it and use information for it to be meaningful. I guess information visualization is in a sense information manipulation because we are thinking about the data and then trying to figure out the next course of action based on what we observe.

Zev Winkelman - Feb 10, 2010 12:21:04 pm

On page 5 the authors state: "...the right representation of a problem, often the right visual representation, can make a problematic decision obvious." They also quote Tufte: "There are right ways and wrong ways to show data; there are displays that reveal the truth and displays that do not."

As an engineer I am inclined to agree, but as a policy analyst I'm inclined to question the definition of the terms "right", "obvious", and "truth".

Has this field really grappled with these terms ?

I think the challenge to Tufte's criticism of the Challenger visualizations is a perfect example of this question.

[add comment]
Personal tools