history

user

Lecture on Sep 28, 2009. (Slides)

Readings

  • Required

    • Chapter 3: The Power of Representation, In Things That Make Us Smart. Norman. (pdf)

    • Chapter 4: Data-Ink and Graphical Redesign, In The Visual Display of Quantitative Information. Tufte.
    • Chapter 5: Chartjunk, In The Visual Display of Quantitative Information. Tufte.
    • Chapter 6: Data-Ink Maximization and Graphical Design, In The Visual Display of Quantitative Information.
  • Optional

    • The representation of numbers. Zhang and Norman. (pdf)

Comments

aallison wrote:

I love Tufte's bombastic style of inducing change through ridicule. I agree with many of his points but think he takes his theory a bit too far. His extreme examples of data-ink maximization seem almost alien to the eye in their minimalism. They seem like the result of design within a vacuum. He does little to show that people better understand the data displayed with his improvements, and design (even visual design) should design with the viewer in mind.

It is also tricky to evaluate his new designs because he does not provide them with much of a context or goal. In my mind, an info graphic should serve the purpose of displaying something interesting about a set of data. Each element chosen should be chosen to emphasize that "interesting something" as well as the data itself.

gankit wrote:

I am a little confused between the idea of reflection wrt. data graphs (in the Norman reading). It seems that Norman is trying to emphasize that data graphs allow the understanding to become better by being more experential and less reflection-based; thus decreasing the cognitive load. But, aren't data graphics really meant to enhance the user's ability to reflect upon the data and make getting insights easier and answers to their questions faster?

Norman writes that, "Reflection requires mental effort, something a sleepy, ill patient is apt to have trouble with". Now, first of all, the patient he just described was completely the opposite. And secondly, Tufte in "The Visual Display ..." (ch 3) emphasizes that we should let go of the doctrine that "Graphics are only for the unsophisticated reader".

Thus, it seems that norman is making a far simpler and highly general claim that the first step to better user engagement (and understandability) is better visualization. But the kinds he is looking at are really simple datasets for trivial tasks. The real power of data graphics, I guess, lies in visualizing complex and large amounts of data to really aid the user in their reflection process!

vad wrote:

The flight representation in Norman (page 10 in the PDF, pages 58-59 in the book) blew me away. Can anyone tell me why I haven't seen it before? I fly a lot, and often buy my tickets from price comparison sites and every time I do a search for a flight I get back a list of six or seven flight which I then have to sift through to find out how exactly they want me to fly. A friend of mine agreed to a 17 hour layover because when she booked a flight she mistook AM for PM (so that is *one letter* difference which equates to 12 hours of life wasted in an airport). On the same subject does anyone know where flight comparison sites to get their back-end flight data from? I would love to implement this component and extend it with interactivity.

vad wrote:

The flight representation in Norman (page 10 in the PDF, pages 58-59 in the book) blew me away. Can anyone tell me why I haven't seen it before? I fly a lot, and often buy my tickets from price comparison sites and every time I do a search for a flight I get back a list of six or seven flight which I then have to sift through to find out how exactly they want me to fly. A friend of mine agreed to a 17 hour layover because when she booked a flight she mistook AM for PM (so that is *one letter* difference which equates to 12 hours of life wasted in an airport). On the same subject does anyone know where flight comparison sites to get their back-end flight data from? I would love to implement this component and extend it with interactivity.

vad wrote:

(removed duplicate)

vad wrote:

Comments do not seem to work.

gankit wrote:

I am a little confused between the idea of reflection wrt. data graphs (in the Norman reading). Although I completely agree with Norman's idea of reducing user's cognitive load in general, I don't think we need to make them more experential and less reflection-based. Aren't data graphics really meant to enhance the user's ability to reflect upon the data and make getting insights easier and answers to their questions faster?

Norman writes that, "Reflection requires mental effort, something a sleepy, ill patient is apt to have trouble with". Now, first of all, the patient he just described was completely the opposite. And secondly, Tufte in "The Visual Display ..." (ch 3) emphasizes that we should let go of the doctrine that "Graphics are only for the unsophisticated reader".

Thus, it seems that norman is making a far simpler and highly general claim that the first step to better user engagement (and understandability) is better visualization. But the kinds he is looking at are really simple datasets for trivial tasks. The real power of data graphics, I guess, lies in visualizing complex and large amounts of data to really aid the user in their reflection process!

wchoi25 wrote:

Norman's paper uses many examples to illustrate what seems to be a simple but important point: different situations call for different representations, some better suited than others. The introduction to the paper does a great job illustrating how ubiquitous and ever-present representation tasks are. Because we live in a world of representations, it is not unusual to be always dealing with representations of representations as illustrated by the short sketch with Socrates.

The Tufte reading was again quite informative, and I thought it was interesting that he tries to formulate and calculate actual "data-ink ratio." I am not sure if this is a useful measure beyond illustrating that some ink used may not be illustrating data. It seems that some ink might be deemed "redundant" but still be useful in making a visualization clearer. As long as it clarifies not blurs whatever fact the visualization is trying to communicate, I don't think the extra ink is harmful.

I found it most difficult to agree with Tufte's transformation of the box plot using his theory (p.123-125). He does not discuss what I think is a crucial feature of a boxplot: it visually communicates that the boxed region is where data is centered as opposed to the two line tails by use of thicker, 2D area. Tufte's redesign entails no such visual cues. While there are situations where this redesign is more useful -- such as a researcher trying to quickly jot down a graph to see obvious trends -- I think a less "data-ink efficient" design also has a place in other circumstances.

wchoi25 wrote:
wchoi25 wrote:

??

wchoi25 wrote:
vagrant wrote:

Halfway through chapter 3 on The Power of Representation, I started reflecting on the notation I've been seeing in some of my other classes--from relational algebra to lambda calculus--as well as the representations I saw in my job of an electrical engineering nature. Certainly, after surmounting a learning curve or two, I could understand the symbols and notation in these schemas, and even use them to solve certain problems, but to someone without a firm background in engineering principles, it would be nigh impossible to explain these representations to them. In other words, I have come to wonder if there exists some hierarchy of representations. Some are simple and easy to grasp for the common man, but are restricted in their computational or flex power. However, perhaps by grasping some simpler representations, one may then graduate to more complicated ones, those that require some prerequisite comprehension, but are more powerful in their controlled capabilities—this may describe some relations between computer science and mathematics, for instance. In some ways, the tic-tac-toe versus 15 example from the chapter alludes to this, albeit from a perspective contrasting human cognition and computer logic. As for the rest of the chapter, I get the feeling that Norman completed this writing well before the advent of Oracle or Adobe.

Tufte’s concept of data-ink is one with which I empathize; from a cerebral viewpoint, I have also always had a vague sense that the amount of ink used in a visualization of data should be minimized. I question if Tufte goes a bit too far in his determination to eliminate unnecessary ink, however. While he does note in his chapter on graphical design that ultimately, some extraneous ink may be helpful from an aesthetic perspective, I find some of his critiques so minimalist that the results would require me to stare at the them for an extra few seconds to comprehend, and even then with only a half bit of certainty. Worse, I believe that for the purposes of repeated reference, each time I was to come back to those critiques, I would have to repeatedly spend that extra time to grasp the information the visualizations are meant to convey. When the iPhone was announced, many mobile phone companies pointed out that the interface’s design required more button presses to perform basic tasks, compared to existing phones with simpler (i.e. less attractive) interfaces. But today, I think most would agree with the sentiment that Apple’s presentation has the better balance of “minimalist” aesthetics and utility. That said, on the whole I agree with Tufte’s criticism of flowery graphics. It is a notion akin to the feelings I get when an essay is presented in an overly-creative font, when a song features too many drum solos, or when a video game sprinkles too many cutscenes throughout its gameplay.

vagrant wrote:

Repeated comment, removed.

nmarrocc wrote:

Norman states that Socrates was against books because he feared the reader would simply absorb the books information without being able to question, think for themselves or have any kind of dialog. Its funny that today, many say the same about TV or video games. How often do we watch something and just turn off our minds? Maybe more often than we think. I would love to see how much of peoples subconscious actions are determined by what they see on TV. Clearly it must have a profound impact on peoples thoughts otherwise there would be no money in advertising. A good graphic can get the information into our heads for sure, but I thinks its still our responsibility to be aware of the influence its having on us. Maybe we could design visualizations that help people see two distinct sides of an issue vs just telling one story...

vagrant wrote:

Repeated comment, removed.

jieun5 wrote:

The most shocking example I encountered in the readings, that conveys the importance of the choice of representation of a given problem, was the tic-tac-toe example (Norman, p. 53-55).

First surprise came to me when I realized that 'the game of 15' could be represented as a very simple tic-tac-toe game; that the winning strategy on the latter can be mapped precisely as that of the former-- as the author calls it "problem isomorphos."

Second interesting insight was the author's poignant claim on how "although the spatial representation of tickatcktoe is much easier for people to play than the arithmetic one of 15, for computers the arithmetic representation is much easier."

I pondered about these two issues, and thought about the pros and cons of the two representations, and to what extent they are "isomorphic" and "computationally efficient".

As for my first surprise, I realized that the isomorphism has to do with the optimal move, and *not* with the symbols that are used to represent the two different games. Clearly, the tic-tac-toe makes a distinction only between the two *players* using the symbols Os and Xs, while the game-of-15 keeps track of the choice of *numbers* using 9 arabic numerical symbols. The most difficult component of problem-solving that is cleverly hidden from view in the tic-tac-toe model is the process of determining the 8 sets of 3 numbers that add up to 15, and arranging them on a 3x3 grid. In other words, if we did not already have this mapping of 3x3 spatial position to 9 numbers (as shown in the middle of p. 54), the tic-tac-toe method would not really save us time.

But provided that we already have an arrangement of 9 numbers on a 3x3 grid (i.e. the mapping scheme between the two representations), then the tic-tac-toe method would be an easier representation of the original problem for humans, precisely because it leverages off of our quick visual-processing of the 3x3 board. But it is unfair to blame the computer for its slowness in the perceptual processing of this "representing world" (using the def. on p. 49), because such visual representation only slows it down.

In other words, the part that is most difficult for us (determining the proper 3x3 arrangement of 9 numbers) is the easiest for the computer; the part that is the easiest for us (seeing a straight line of patterns in the "representing world") is the most difficult part (and an unnecessary step for getting the correct answer in the "represented world") for the computer.

dmac wrote:

Chapter 6 in Tufte was interesting because it was the first real place that he developed improved methods for visualization built from established methods and the principles laid down in previous chapters. However, in a few of his designs I think he went too far to make a point. For example, while he claims the improved off-by-one-pixel quartile plot (page 124) is easy to read, I have a hard time gauging where the quartiles and median are simply by glancing at the line. Similarly, I found the rugplot (page 135) more confusing than enlightening, especially lacking an idea of what sort of transformations Tufte imagines would be used to create this visualization. Finally, while I did like the idea of the range-frame (page 130), we discussed in lecture that when plotting quantitative data having the zero visible is important, and using a range-frame often eliminates it. Later in the text Tufte puts a disclaimer saying that the readability of a visualization should be placed over any of his principles, and I think in parts of this chapter he sacrificed that for the sake of showing off some new ideas.

zdevito wrote:

One common decision for all of these plots was the choice to use a linear or logarithmic scale for the MIC of each antibiotic. Since the data spanned 5 orders of magnitude, those who represented the data linearly needed to use techniques such as breaking the scale, or mapping the values to the curvature of a circle in order to include all the values linearly. Those who choose a logarithmic scale could easily fit the data spatially but had to be careful about what to represent as zero and how to make it obvious that the scale was logarithmic. In fact, we discussed how the audience may affect his choice -- a non-technical audience may have trouble understanding logarithms naturally. According to an article published last year in Science (http://www.sciencemag.org/cgi/content/abstract/320/5880/1217), it is more likely that we are born able to naturally interpret the logarithmic scale, and that the linear scale is learned as a result of our number systems and understanding addition/subtraction. Researches studied both western subjects and those from a Amazonian group with a simplified number system and little formal education. Given tasks to map quantities (numbers of dot, tones, numbers written in language) to a place on a line, those from the Amazonian group would map the values more logarithmically than linearly (e.g. 3 is at the center of 1 and 9), while those from a western background would map them linearly (5 is at the center). This result seems relevant to data visualization: it points out that some of the mappings that seem 'natural' to use as visualizations like the linear scale may be learned. This seems to suggest that we can learn new visual encodings to the point where they become intuitive to us even if they are not necessarily inborn.

fxchen wrote:

Earlier today I reflected on these readings again in context of Assignment 1 (MIC for certain bacteria). I'm starting to realize both the power and the thought process a careful visualization creator must follow. I unfortunately made some rookie mistakes; I'd bet Mr. Tufte would not be happy with me.

Improvements: Encodings were adopted to represent data disparately, but never really tie together to create a narrative. Some of visualizations crafted a specific analysis using the decreasing trend in penicilin and the gram staining. Making the assumption each bacteria was unrelated to the dosages other bacteria were treated with, I created a vis. focusing on the best antibiotic to combat this bacteria. I'll definitely focus more on finding a trend or some sort of strong correlation or create a narrative instead of simply better representing data in future assignments.

codeb87 wrote:

I liked Norman's discussion of choosing visual encodings that we perceive in a similar way to the way we perceive data itself (ie color density/saturation is a good choice for additive data, like percentages). I'm interested in how we could extend this principle into more exotic domains that go beyond nominative, ordinal, and quantitative. Particularly it made me think about music - what medium could we use to depict different chords of a song? Is there perceptual psychology related to color theory that could link well with a musician's knowledge of how different chords make us "feel"? Also, could we map the different timbres of a song to different textures on a graphic display? To what degree could we make these design decisions scientific and deterministic rather than artistic? I think it would be interesting to pair perceptual music theory with our current discussion of how effective different visual encodings are. Moreover, I want to know how different visual representations can make us FEEL. Obviously this could be expanded to data besides music, but this particular topic really got my brain juices flowing.

jqle09 wrote:

Tufte's discussion of chartjunk is possibly one of the funniest chapters of any book I've ever read. You can really feel the disdain he holds for chartjunk and its proponents. I think at times he sounds a bit overzealous in his criticism of chartjunk, but his main point that designs need to be "intriguing and curiosity provoking, drawing the viewer into the wonder of data ..." is important.

The chapter on data ink made me think that Tufte had too many run ins with the wrong ink provider leading to his enthusiasm for saving as much ink as possible for data. I think he goes a little too far with his redesign of the bar chart on page 102 but overall I think his point, that the data must be shown, still stands.

Though I felt that some of Tufte's proposed redesigns were not as clear as he would have hoped (the single line box and whisker plots are particularly hard to understand and read). It might be that I am just not familiar with these styles, and familiarity allows for quicker comprehension.

Most of all Tufte's chapters had me thinking about ways I could have greatly improved my assignment 1 visualization. There were many stories that could have been told with that data and the narratives for all of them could have easily, but most importantly I think clearly, been contained in one graphic.

I think Norman's chapter really reiterated a lot of Tufte's ideas even though his goal was to describe what makes humans smart. He states that a "good representation captures the essential elements of events ... " providing substantive power to enhance people's ability to think. I think there is also a feedback loop here, where we create representations of our abstractions that amplify our cognition with regard to these abstractions and further let us improve them so later representations also improve, and so on. It seems thinking in this way is self reinforcing and works very well in improving thought, and making us smarter.

In response to a comment above about musical perception, I think an interesting experiment might be to have a progression of notes the person hears and then have them sing what they think the next note in the progression should be. I would want to hear the person sing that note because I think that might give us insight into how they encode musical information. I then think it would be interesting to see how what notes are picked varies across different cultures. I'm thinking this might be similar to this, http://blog.doloreslabs.com/topics/colors/, color experiment. Though there would definitely need to be some way to correct for varying levels of musicality. Though watching this video, http://www.youtube.com/watch?v=ne6tB2KiZuk , made me think that wouldn't be a problem.

alai24 wrote:

I agree that Tufte takes the data-ink minimalization a bit too far in the bar plot example. Unless it was being printed in gold I see no advantage to using the half bars instead of a full bar. Perhaps it is because I believe the regular bars are more intuitive after a decade of bar-plot indoctrination, so some others may think differently.

The point made on avoiding the use of pattens in datagraphs is well made. These cross hatching and moire vibrations seem to have a predilection for causing optical illusions (like oiangledlineswaves) so I can see why they should be avoided.

Norman's discussion of representing numbers was pretty interesting; I was surprised to see how easy it was to add using Roman Numerals but I can't even fathom whatd it'd be like to try to multiply them. Binary numbers are pretty easy to add and multiply by hand - maybe we should switch!

nornaun wrote:

Before I start rambling, is the wiki a little delayed? I have checked this page super regularly and saw no comment at all until just now (3:21AM Monday morning) when I saw that there have been posted commented since 9/25. :S

I am a fan of Norman, so you can say that I'm biased. I like the way he makes distinction between experiential artifacts and reflective artifacts. It opens a new perspective in thinking about how data should be represented to accomplish different tasks. This can be linked back to Card and Mackinlay's paper, the Structure of the Information Visualization Design Space, which says that visualizations have two main uses: to communicate and to solve the problems. To simplify the concept, I can say map: experiental -> communication and reflective -> solving problems. Life is not as simple as it looks and there are special cases but this is the picture I get from the two readings.

One point that I want to raise is about Tufte's data ink density. On one hand, I think it is a concrete way to measure the "efficiency" of a graphics. On the other hand, I still believe that a graphics is not only for data derivation. In "Emotional Design," Norman discusses three different aspects of design:

  • Visceral: concern with appearances Behavioral: concern with pleasure and effectiveness of use Reflective: concern with self-image, personal satisfaction, memories

I would argue that Tufte focuses on visceral and behavioral aspects of presentation. However, the reflective part, which is also important, are neglected. Why is reflective aspects of graphics is important? One argument is to create incentive. Although I am not an elementary school students, I still love to look at pretty graphics. What is the point of creating high data-ink density graphics if it puts stones in the hearts of its viewer? I believe that data visualization can achieve the same effects as products designed for reflective design. (For example, interactive graphics that can change colors & themes to fit a viewer's preferences?)

If I am an ideal human being who always think rationally, I might totally agree with Tufte's suggestion. It is efficient and effective. Alas, I am not. For me, some eye candies--to motivate me to look at the graphics and to stimulate my thought--doesn't seem so bad.

bowenli wrote:

@gankit, Yes I agree with you in that I think the definition of "reflection" may be the point of disagreement. I think Norman meant the point between actively deciphering a set of data vs. gaining some more intuitive understanding of it. Though the latter is not necessarily the lack of thinking, rather than aiding it.

The Representation of Numbers: Wow, what an in depth look at numbers. I think the point about doing algebra and math manipulations in different systems is really interesting. And it potentially shows why the Arabic system is so popular in the modern world - where we have to use much more complex math than the ancient times.

@vagrant Regarding the heiarchy of representations. I think this may also have to do with a concept introduced in the Numbers paper - regarding what type of information is "external" meaning it is apparent by looking at it, and what types are "internal" which is semantic meaning given your understanding of the symbol, but not inherent to it. I think perhaps the higher level symbols you suggest do exist and would contain more of these internal representation type dimensions, thus being more coginitively difficult to use but are ultimately more complex.

akothari wrote:

From Norman's reading, I really enjoyed the discussion on the airline guide - it's interesting how much the mind is aided by different representations. I wonder if big airline companies such as American Airlines and United Airways have read these reports. I tried their websites a few months ago, and it was an awful experience. Bad enough to pay $50 more and fly Virgin America.

I was less impressed by the medical prescription example. Although it does pack a lot of information in a clean table, specifics are being lost - for example, what if there are two tablets, or what if the time is really important (take tablet before meals vs. night). Overall, I thought it was a great read.

cabryant wrote:

Although Tufte may be correct that bilateral symmetry constitutes redundancy, his argument that symmetric portions should be removed from face diagrams (p. 97) does not necessarily follow. His reference to studies indicating that subjects thoroughly examine a symmetric half and only briefly verify parity in the other actually acknowledges the conceptual need for symmetry. More favorable evidence in his defense would be studies indicating that, when presented with partially occluded images, subjects perform perceptual completion in accordance with theories of expectation-maximization. Nonetheless, such behavior likely coincides with additional cognitive effort that is being diverted from the task of deciphering the information in the visualization. Furthermore, as in the case of “half-faces,” higher order cognitive tasks may also be introduced, wherein the viewer might begin to question the semantic significance of non-occluded half-faces (symbols which violate expectations of visual processing). If this occurs, then the marginal benefit derived from removing symmetric information may be less than the ambiguity and excess processing that arises due to violations of expectation.

On an unrelated note, I was struck with the profundity of Norman's claim that “things not represented fall in importance: They tend to be forgotten or, even if remembered, given little weight … We value what we can measure (or represent)” (pp. 52-53). Similarly, we tend to value ways in which we can represent measurements over ways that we should represent them, leading to the “We-Used-A-Computer-To-Build-A-Duck[-Because-We-Can]” syndrome described by Tufte. These dangers apply to fields beyond data visualization, including my current course of study: the design of technologies that enhance learning. History provides numerous examples of the promotion of alluring technology that is inappropriate for the target learning space, and may even constitute a detriment when compared to fundamental tools such as pencil and paper. It is a disservice to any field to promote potentially distracting complexity over appropriate simplicity.

rmnoon wrote:

I generally really like Tufte, but one thing he does that makes me cringe a little is his evisceration of box and whisker plots (like on page 123/124 and 129 in TVDOQI). Stripping away ink is great, but conveying the most important information (the median/IQR) by subtle variations in line position or weight just seems like folley. I'd look at those plots and see a printer's error, not important data. Tufte is such an ideologue sometimes.

Also, Norman's discussion of how different representations can make us smarter reminds me of visual programming languages (http://en.wikipedia.org/wiki/Visual_programming_language). I've never really used them, but I've heard that even young children and other non-traditional programmers can do amazingly complicated stuff when the representation of control flow and logic is transferred out of the 2d document representation. Does anyone have any first-hand experience with this?

rajsat wrote:

The Norman reading talks a lot about how a good representation helps in capturing the critical features of the represented world and how it enhances the process of interpretation and the ability to make judgments & decisions. Visualization of music is something that has always interested me. I like the way some music visualizations bring out the essence of emotions and feelings that are inherently part of the song and gets you deeply rooted into it. One primary focus of music visualization would be to identify a description of music formal enough that it can be considered as the description of an image. But then the question arises- what description gives us the image that we want. There are a million possibilities- the complicated ones take even emotions brought out by music into account, but most of the general visualizations that we see on common software like iTunes take basic description like harmonics, tempo, frequency into account. For instance, I have always felt (ofcourse many of you might choose to disagree) that the pleasurable(satisfying experience) that I get while watching the visualization while listening to the music is considerably more than listening alone. There are somethings the visualization can just not achieve - like words(lyrics) -that's something thats hard to code in visual form and maybe other intricate stuff like rhythm,chord change/progression, etc. I stumbled upon this paper while Googling about music visualizations. Though its core is based mostly on mathematical stuff that I'm still oblivious about, the essence is to develop a music player that synchronizes visualization/photos with music - the basis of which is the emotions that are evoked by auditory stimulus(from the music) and the visual content (in the photo).

joeld wrote:

The opening passage of Norman is quite memorable, but I think his best example is the description of the game called "15" which he later reveals to be tic tac toe. Right now I am working on a data understanding project with lots of samples and possibly hundreds of metrics that people could be interested in and I keep hoping that I'll come across a visualization that will make it seem like it's just a simple to understand game of tic tac toe.

That said - the visual representation of tic tac toe as a game elucidates the rules of the game in a way that a child can understand and begin to play. The spatial nature of the game, symetry and the need to perform some blocking moves become apparent quite quickly. What is still missing from this representation is a representation of the "state-space" of the game. A "simple" graph like this would allow one to compute the optimal move in either game as well as allow one to estimate the probability of winning.

I use simple in quotes here because the visual representation require a few hundred nodes if it were a graph. At this level of detail one begins to wonder if the best "visualization" is a look up table.

malee wrote:

I disagree with Tufte's reading of Venturi et al's "Learning from Las Vegas" (Page 117), which I had to read for an architectural history class last year. I agree with Tufte when he interprets the "duck" as a creation where structure is synonymous with ornament and decoration. As in the picture of the duck-form store on Page 117, the underlying structure of the building is hidden and deemphasized with respect to the overall message and symbolic form of the duck.

However, I disagree with Tufte when he tries to use the notion of the "duck" to make a point about chartjunk. A "duck" is a building that conveys one message with its entire form. Thus, a "duck" in architecture is quite clear in conveying a point. Each part of the structure is meant to add towards this message, without distracting the viewer. Similarly, each part of a graphic should add towards the overall point, and its underlying structure (such as gridlines, axes, etc) should be deemphasized to highlight the main point. In this way, I feel like a proper interpretation of Venturi et al is that "ducks" are actually good examples of design. On another note, I felt sort of ridiculous writing about ducks as buildings in my architecture class, and it sounds even sillier in a CS class. :)

tessaro wrote:

Given Norman's distinction between experiential and reflective representations, it would be interesting to hear his views on graphical hybrids like tag clouds. (For clarity, in the unlikely case that the term is unfamiliar, a 'tag cloud' organizes text information so that words of greater frequency in say an article or blog are scaled to a larger point size while other less frequently used terms are shown smaller. The aggregate image is shown in a 'cloud' like array. The scales usually have a baseline size for legibility as well as a maximum upper bound for the most frequent.) 

Does the reflective decoding of text coupled with natural spatial encoding in a tag cloud support or undermine Norman's Naturalness and Perceptual principles? Would such representations fall squarely between the graphic encoding of say a bar graph with labels and that of a simple ranking, like an ordered list, in terms of timed performance? Is there a delay in decoding the text content or adjusting for the common issue of judging word length as another indicator of fequency (as opposed to point size alone)?

My guess is that utility of the tag cloud to indicate at a glance is in opposition to studied reflection. It cannot indicate value distinctions of the small range between 284 to 312, as in Norman's example. The value of a tag cloud as perceptual mapping is I think as a trend indicator or pointer to large differences. It can also indicate general equivalence or low variation, if all the words are roughly the same point size, but that seems tangential to its intent. ( The area encoding of a tree-map would be better suited to the task of making finer distinctions through true area mapping with perhaps the addition of a gradient.)

The perceptual efficiency of size is found not by explicitly quantifying nuanced distinction per se but by speedily showing significant difference as well as the mere presence of slight variation as something noteworthy. Tag clouds can function something like information portraits in this regard but fall short when the ambiguity of an ordering based on text size alone asks more questions than its encoding strategy can possibly illuminate.

saip wrote:

One of the more fundamental points to be driven in from the discussions, both in Norman's readings and from the in-class presentations, is that any visualization is deemed more or less apt depending on the audience it is intended for. This plays the major role in governing the type of visualization and the amount/type of data to be conveyed. From a more philosophical point of view, the "Paradox of Choice" should be considered while deciding on the visualization, while keeping the intended audience in mind (http://en.wikipedia.org/wiki/The_Paradox_of_Choice). A gist of the paradox is that More is Less. The more important information of the lot is to be represented in the most comprehensible form; representing, or trying to pack in, all the available in a visualization often ends up conveying less to the reader, mainly due to information overload and the cognitive inability to make intuitive sense of too much.

For instance, the graphical representation of flight timings (which was awesome by the way) offers little benefit in conveying all the information that the text-table does. Also, the drug prescription example from Norman's; while ordering by drugs is more confusing for a patient, this seems to be the most convenient from the point of view of a drug store owner. In both cases, multiple visualizations might have to be provided, depending on the audience it is intended for. Each addresses a different problem/audience and provides the more important sector of information for that problem/audience. This fact was also very evident from the in-class presentations of various visualizations of the antibiotic effectiveness. While there were several fabulous visualizations, each meant to convey a different kind of information, intended for possibly different groups of people.

anuraag wrote:

Tufte's data-ink maximization chapter was an interesting read, but, like some of the other commenters, I found a few of the re-designs jarring. In particular, the line offset approach he seems to favor for box plots and chart axes seemed an odd choice. Tufte argues "It uses the ink effectively" and asserts that it "looks good," but he offers maddeningly few justifications for this.

I thought the most intriguing insight in the chapter was when Tufte notes that "the large reduction in the amount of drawing is relevant for the use of such designs in informal, exploratory data analysis, where the research worker's time should be devoted to matters other than drawing lines." Saving time for quick-and-dirty visualization is a noble goal, but I wonder if Tufte's re-designs actually achieve this. Doing things like off-setting a line by a small amount probably requires more time than creating a traditional box plot, because even though a box plot requires more ink, it requires less precision. Carefully drawing small offsets is hardly likely to save a research worker's time. Similarly, the data-ink maximization techniques that involve erasure don't match this time-saving principle -- in this case, having less ink requires more steps!

saip wrote:

One of the more fundamental points to be driven in from the discussions, both in Norman's readings and from the in-class presentations, is that any visualization is deemed more or less apt depending on the audience it is intended for. This plays the major role in governing the type of visualization and the amount/type of data to be conveyed. From a more philosophical point of view, the "Paradox of Choice" should be considered while deciding on the visualization, while keeping the intended audience in mind (http://en.wikipedia.org/wiki/The_Paradox_of_Choice). A gist of the paradox is that More is Less. The more important information of the lot is to be represented in the most comprehensible form; representing, or trying to pack in, all the available in a visualization often ends up conveying less to the reader, mainly due to information overload and the cognitive inability to make intuitive sense of too much.

For instance, the graphical representation of flight timings (which was awesome by the way) offers little benefit in conveying all the information that the text-table does. Also, the drug prescription example from Norman's; while ordering by drugs is more confusing for a patient, this seems to be the most convenient from the point of view of a drug store owner. In both cases, multiple visualizations might have to be provided, depending on the audience it is intended for. Each addresses a different problem/audience and provides the more important sector of information for that problem/audience. This fact was also very evident from the in-class presentations of various visualizations of the antibiotic effectiveness. While there were several fabulous visualizations, each meant to convey a different kind of information, intended for possibly different groups of people.

rnarayan wrote:

I agree with several comments about Tufte's offsetting quartile plot - unless attention is explicitly called out to the quartile with additional text, it is barely noticeable. I wonder if Tufte's arguments about data-ink maximization is a bit antiquated, since his observations are before the advent of hi-res computer graphic displays. Another example is the space consumed by bilateral symmetry and the ease of sorting - again it seems his comments could be only relevant for 1980s when the book was first written.

In a somewhat related note, I found an interesting data-rich website (policymap.com), which crams over 10K indicators in a geo-spatial information space (it was in the news last week, since its mapping software provider was acquired by Apple). The density of data here is overpowering, although the interface could be structured quite differently - it seems to me that such problems are a lot more relevant in today's info-exploded context, than Tufte's points about data-ink maximization.

Leave a comment