Notes from Tufte's seminar, Presenting Data and Information
A.J. Bernheim Brush
  [home]
 

Tufte's seminar, aptly named Presenting Data and Information, focused on how to improve your presentation of data and information. Tufte used examples from his three books, The Visual Display of Quantitative Information, Envisioning Information and Visual Explanations to motivate each point he made.  During the seminar he listed the principles of information design, gave a tour of his new book Visual Explanations, discussed how to display financial data, and outlined ways to become a better presenter. He also spoke about his thoughts on computer interface design and used the Challenger accident to highlight the importance of displaying data to show the truth.

A recurring theme throughout the seminar was the importance of good content and good presentation of that content. You want your viewer to puzzle over your content, rather than your design.  In fact Tufte said "often the role of a good presentation is just not to screw up good content."  He suggested putting together a portfolio of 10-15 "insanely great" information displays in your particular field.  Then when you need to create an information display you can look at the portfolio and determine "which architecture to plagiarize today."  He quoted T.S. Elliot who said "Talent imitates, Genius steals."

During his presentation Tufte used no overheads or slides, instead referring us to figures in his books. To get the most out of these notes you will want to be looking at the books when you read them.   (Yes, I am violating Principle of Information Design #4.) Tufte also gave us each a reading list of recommended books for information design, which I would be happy to copy for anyone who is interested.  Although Tufte only briefly spoke specifically about computer interface design,  he highlighted the fact that the principles of information design are consistent across varying presentation technologies.
 

Tufte began the seminar with his two major problems in the display of information:

1. Everything interesting is multivariate (three or more dimensions of information) and we only have a two dimensional display surface.   He termed this the "Escaping Flatland Problem"

2. How can you improve the data transfer of your presentation? We want more information to pass from you to your listener.
 

Principles of Information Design

Next, Tufte illustrated his principles of information design, using Minard's graphic showing the fate of Napoleon's army in Russia [Quantitative Information, pg. 41].  These principles help deal with his two major problems of information display.

1. Enforce visual comparisons.
Answer the question, compared with what?  Minard shows the varying size of the army throughout the campaign.

2.  Show causality, the mechanism, and process.
Minard tied the decreasing size of the retreating army to a graph of the brutally cold temperatures.

3. Display multivariate information.
The world we want to understand is multivariate, anything with a small number of variables probably isn't that interesting.  In Minard's graphic we see 6 variables: the army size, location (x,y), direction of travel, temperature and the date.

4. Integrate text, numbers, images, into a complete picture.
Don't make your presentation suffer from mode of production, with figures attached at the end etc.  Keep your argument integrated.  Minard does this in a number of ways,  including direct labeling of the army's location right on the map, and integrated captions.

5. Presentations stand or fall based on the quality, relevance and integrity of your content.
Minard was obviously a talented engineer, in addition he produced his graphic as an anti-war poster and thus cared deeply about the content.

6.  Show information adjacent in space rather than stacked in time.
For this point we looked at several pictures, including illustrations of the sun spots on Saturn's by Christopher Scheiner, [Envisioning Information, page 19].  Here, several small pictures of Saturn are shown next to each other so that you can easily see the change.  Another illustration by Repton, [Visual Explanations, pg. 16] shows the difficulty of comparing two images stacked in time (one behind the other).

Tufte tied this point to computer interfaces, saying that computer interfaces show information stacked in time as you proceed sequentially from one screen to the next.  Often people need or want to compare the current visible state to the previous visible state, but can't access the previous state.   He suggested we all go out and buy great big high resolution monitors that would allow more comparisons within the eye span.

7. Use small multiples.
 Scheiner's 38 little Saturn images [Envisioning Information, pg. 19] support the advantages of small multiples.  With small multiples you can escape the problem of only two dimensions and also index on time.  Small multiples have high resolution, they are adjacent in space and easy on the viewer.  Once the viewer has learned how to read one picture they can use that knowledge for all the remaining images.  Tufte also pointed out that small multiples seem to have "inherent credibility" with readers, which he felt might come from the fact you are showing the whole dataset and also your mastery of detail.

Examples of small multiples pervade the three books, but another one we looked at in particular was in a novel medical interface proposed by Tufte and Powsner [Visual Explanations, page 111].
 
 

The Meta-Principle of Information Design

Tufte also presented his meta-principle of information design.   Basically the principles that are important arranging design displays are the same as in analytical thinking.  You should ask yourself, "What's the task and how can we support that?"  He emphasized that the principles of information design are important regardless if the display technology is the web or a map carved into a rock.
 
 

A Tour of Visual Explanations (his new book)

Here Tufte went through each chapter in his new book highlighting things of particular interest.

Chapter 1. Images and Quantities
Watch for reshaping of content by form.  Don't overwhelm your data with a dark grid, as the Center for Supercomputing did in their videotape "Study of a Numerically Modeled Severe Storm" [pg. 20].

Chapter 2. Visual and Statistical Thinking
This chapter looks at the importance in presenting data truthfully so that the correct conclusions can be reached.  The examples used are  the Challenger accident [pg. 39-53] and John Snow's detection of the cause of a cholera epidemic [pg. 27-37].

Chapter 3.  Explaining Magic
Learn from magic what not to do.  Magicians are trying to make their presentation hard to follow, so let's do the opposite of what they do.
For example:
Magic: Don't tell about the trick in advance
Presentation: Tell the audience the problem and who cares right up front.

Chapter 4. The Smallest Effective Difference.
Make your visual movement as minimal as possible, but absolutely clear so that there is a just noticeable difference.  A redrawn diagram of the ear [pg. 74], and a redrawn map [pg. 76,77] demonstrate the advantages of this technique.

Chapter 5. Parallelism
Some Repton architectural drawings [pg. 80,81] are shown both stacked on top of each other and side-by-side to demonstrate the advantage of showing pictures side-by-side for comparison,

Chapter 6. Multiples in Space and Time
An entire chapter of small multiple pictures.

Chapter 7. Visual Confections
Examples (including one with Babar the elephant) of confections.  In a confection many visual events are selected and placed together to tell a story.  This chapter also includes an example of his computer interface for a museum guide [pg. 146,147].
 
 

Presenting Financial Data

Tufte's first book, The Visual Display of Quantitative Information focuses on pictures of numbers.  He summed up the major points of the book for us.

1. Show an assessment of change. The graphic of Connecticut Traffic Deaths is a perfect example of misleading the user by not showing an assessment of the change [Quantitative Information pg. 74].

2. Understand the average and variability of the data. The micro/macro detail of the New York Times temperature graphic shows both the details and overall trends [Quantitative Information pg. 30].

3. Clutter and confusion are evidence of failure of design.

4. Don't trust a display if it doesn't have footnotes or source information.
Always included documentation about where the data comes from.

5. Data requires adjustment and massaging before presentation.
For example, you must adjust for inflation if you are comparing dollar amounts for more than a few years.  The graphic of the Total Retail sales in the US shows all the adjustments necessary before the retail sales data is actually meaningful [Quantitative Information pg. 38].

6. Move from particular to more general.
 Tufte used a disturbing graphic, [Envisioning Information, pg. 56,57] that shows the cost of a treating one health care patient, to illustrate macro annotation on micro detail. The final footnote on page 57 is a good example of moving from particular to general
 


Thoughts on Computer Interfaces

Tufte discusses computer interface, in Visual Explanations starting on page 146.  On these pages he presents his interface for a museum guide.   He advocates a flat interface where all possible options are visible at once.  He also stresses that a minimum amount of the screen should be allocated to what he calls "administrative debris," and shows one interface where only 18% of the screen is used for content.   Tufte quoted Cooper who said "No matter how beautiful your interface is, it would be better if there was less of it."

For web site design in particular, Tufte stressed that the only thing unique and interesting about your web space was the content (not the rotating logo). People who come to your web site need to know immediately what they can learn from it.  He also discounted a current belief in web design that you should only put 5-7 things on a page.   This belief is mistakenly based on research that proved people have trouble memorizing more that 5 to 7 unrelated things.  However,  people visiting your web site are not trying to memorize information, they are scanning related things for something they are interested in.


Challenger Accident

After lunch we looked in depth at the Challenger accident as an example of what can happen when data is presented in a bad way.   We examined the charts rocket engineers produced to try and convince NASA not to launch the Challenger.  The story and failing of the graphics are fascinating and I urge you all to look at pages 40 - 53 in Visual Explanations.
 


Ways To Become a Better Presenter

Tufte closed with a list of 15 ways to become a better presenter.

1. Show up early (hand out stuff, say hello, find out if room is overbooked..)

2. Immediately state the problem, who cares about this problem, and a possible solution.
  - Never apologize for yourself (because you have a cold,...)
  - Stay out of the first person, concentrate on the content.

3. Explain complex point using the Particular, General, Particular Technique.
 - First a Particular detail of graphic (that 13% means...)
 - Then give General overview and context of graphic
 - Finish with another Particularly important detail

4. Give everyone in audience at least one piece of paper.

5. Think about the audience by what they read (NY Times, Wall Street Journal...) They are just as smart when listening to your talk as when reading.

6. Avoid overheads.  Overheads lead to stupid bullet lists and encourage sloppy generic thinking.  Avoid powerpoint.

7. Respect your audience as colleagues.  Assume the people in audience care as much about your content as you do and that they are as smart as you are.

8. Humor is a valuable tool. It allows for hyperbole and exaggeration. But you must be careful,  and use humor to reinforce a particular point.  Don't insult or alienate your audience.

9. DON'T use male pronoun for general examples.  Go into the plural when speaking.  For example,  say "When the user moves the cursor, they ..."

10. When answering questions, be careful.  Some audiences will judge your entire performance on your response to questions.  When you ask for questions, count silently to 10, allowing people to ask questions.

11. If you believe in the stuff don't stand on the stage clutching the podium.  Let them see your gestures, and honest enthusiasm.

12.  Finish early!  Something good is bound to happen.

13.  Practice, Practice, practice!  In front of other people, and in front of a camera.  Watch your videotape at normal speed and at high speed.  Then turn off the picture and listen to the sound.

14. Beware of dehydration.  Travel and giving talks are very dehydrating, drink lots of water, skip the alcohol.

15. All this advice comes after the content. Most of what happens depends on the content.
 
 
 


ajb@cs.washington.edu
Last modified: 2/07/2000