11/9/99

Seminar on Presenting Data and Information

11/8/99

Notes Regarding

Jeff Moore and Greg Jones and I attended this seminar in Crystal City given by Edward R. Tufte.

2 PROFOUND ISSUES

Everything interesting is MULTIVARIATE i.e. the issue of DIMENSIONALITY

— involving three or more variables

— we live in three space + time and we have to reason about that

— but we do that in a 2D flatland of paper and display screens

— We need to escape from the 2D display and the tyranny of flatland

INFORMATION RESOLUTION is the second issue both in static and dynamic terms

— static, pixel resolution: dynamic, information transfer rate

— exhibit: 1st English Edition of Euclid signed by Ben Jonson (1590)

— The Dreaded Letter Code: example of figures marked A, B, C makes your users into puzzle solvers instead of readers. Generally labels should be at the end of data lines not letter codes.

— Example of three dimensional pop-ups as one way people have used to escape flatland. This is an example of Brute Force i.e. making it 3D

ENFORCE VISUAL COMPARISONS (Compared to what? example of Napolean's Russian campaign.)

I. Wise Comparisons (422K people to 10K survivors)

II. Show Mechanism/Dynamics/Causality — Why did this happen?

Causality thinking ~ descriptive so far

III. Complex World (Higher Dimensionality

ex. temperature scale along bottom of Napolean chart

a total of six variables (size of the army, location in 2D, direction, temperature, dates)

IV. Completely Integrate — text, image, word, map. Don't break up the evidence and give your readers a puzzle to solve.

V. CONTENT — Tell a STORY integrating Quality, Relevance, and Integrity of CONTENT — reasoning about the content is important. If there is a problem it is almost always due to the need for BETTER CONTENT that is 1) On Point, 2) Relevant, and 3) Meaningful. Don't Screw Up the CONTENT!

The Five Principles reflect a METAPRINCIPLE

— Where do ideas about information design come from?

— What are the DEEP PRINCIPLES of Information Design?

Compared with What?

Displays Support Thinking — What is THINKING TASK? How to do it?

WHAT IS THINKING TASK? — How to arrange my data to help out, make thinking easier?

GOOD DESIGN is Visual Thinking i.e. Clear Thinking Made Visible

BAD DESIGN is the reverse, i.e. BAD Thinking.

Beware of CHART JUNK, meaningless throwaway slogans. People who demonstrated that they don't know anything about the data they are presenting don't have anything to say. BEWARE.

Look for the RING OF TRUTH, ex. Galileo's illustrative use of diagrams to show what he saw embedded in the text, Richard Feynman and his illustration during the Challenger disastor.

SHOW INFORMATION Adjacent in Space NOT Stacked In Time.

— stacked in time is just One Darn Thing After Another — the significance of the data is rapidly lost. Relentless Sequentiality and the Question: Where Am I?

If you are showing data on a screen Get a BIG MONITOR

USE SMALL MULTIPLES — illustration the evolution of sunspots taken from a Galileo contemporary — they 1) enforce comparison, 2) provide high resolution, 3) escape flatland (a lot of dimensionality), 4) Easy on the Viewers (learn how to understand one and transfer the understanding to the others immediately by FRAME RECOGNITION, 5) Provide Natural Credibility (as does lots of data anytime -- the Spirit of Knowledge and Truth — the cartiographic (i.e. MAP) metaphor as pictures of NOUNS and VERBS

LINK BETWEEN SEEING and THINKING

Don't Distort the Data -- ex. NASA flyover of Venus, huge vertical distortions, worst mountain is a 3% grade

Square Page => Square Country (illustrated with a published map of England that was distorted because the map would not fit the page otherwise.)

SMALLEST EFFECTIVE DIFFERENCE

Information is about making differences, but you want to unclutter the space by making the visual differences as small as possible while still making the point

- direct labeling

- color metaphoring (Don't saturate the design space -- the DREADED RAINBOW)

- illustration from Newton's Optics of illustrations subsequently distorted and rendered incoherent by separating the figure from its annotation. Separation makes for a disaster for the information. (Newton was shy, so there is only one copy with his name on the title page)

INTEGRATE test/word/iamge

- see cycligram page 95

- visual confections (use of small multiples as ornamentation)

- compartments/frames

- ex. the Potomac River Dangers Illustration, a lot of visually effective information

- Gotti Witness chart -- instance of DEQUANTIFICATION, elimination of unnecessary information from the chart, but always dangerous since it reduces content. Make sure when you dequantify that you are getting what you want.

- Assessment of Change, instance of regression towards the mean -- just because there is a change doesn't mean it is significant. Data Needs To Be Presented In Context -- so the data that forms the context must be presented.

- Provide Location and Variations DETAIL HELPS CREDIBILITY and it is particularly helpful if part of the detail CONFIRMs PRIOR KNOWLEDGE on the part of the viewer.

1. Confusing? — LOOK TO THE DESIGN

2. Money versus Time must always be adjusted for inflation is any significant amount of time (the appearance of improvement is an artifact of a) not applying a deflator and b) not providing comparables)

3. PRICE Indicies as deflators are widely available, all kinds

4. Don't Trust a Display without footnotes (Credibility) PROVIDE FOOTNOTES

5. 90% of Financial Displays are Entirely Descriptive -- What? but not Why?

— Annotation for Why? provides causal explanation

— ex. the Slow Costly Death of Mrs. K, illustrates extensive documentation of a series of hospital bills culminating with the death of the patient.

6. FIND A GREAT DESIGN AND COPY IT — Don't make the user/patron decode a new puzzle. Collect a set of Great Designs and use them as a toolkit. There is a museum of great designs in these three books. To get good design you need programs that can 'see' -- ex. Corel Draw, and Adobe Illustrator (which he particularly liked) — Good Typography requires control of line-idth. You need a set of 10 to 15 'insanely great designs.'

- illustration: data spark preceding a line of fundamental statistics, ex. mutual funds information

- Things to avoid: 1) unfortunate metaphors, 2) kiosk parady, i.e. designs that mirror not content but the hierarchy or bureaucracy of the institution, or 3) worst the deep architecture of computer programming, AVOID DEEP MENUS, A FLAT INTERFACE IS BEST. 4) a lot of people are fleeing web sites -- example of people whose stay at the web site is shorter than the download time of the opening page. What is needed is CONTENT not fancy logos, large pictures, ads, etc.

7. CONTENT not DESIGN -- the design needs to be self-effacing compared with the content. Allocation of SCREEN REAL-ESTATE is the most important and precious resource. Most display are contemptuous of content -- so they don't have any. Instead they exhibit

- Operating System Imperialism (buttons, arrows, spinning gadgets, scrolling frames etc.)

- Dominated by Bad Design Metaphors driven by bureaucracy and marketeers

- MEASURE the screen real-estate devoted to various features. If only 18% is devoted to content you have a bad bad system DEFEND CONTENT

8. Look at HOW THE REAL-ESTATE is Allocated — DOES IT HELP HUMAN UNDERSTANDING? You want to leave TRACES that illustrate the Causality: Why are we having this meeting? -- hand out at least one piece of paper with something meaningful on it.

9. Don't Get It Original -- GET IT RIGHT

- illustration of the patient monitoring system with large numbers of small multiples arranged in vertical columns by major pathology, all on one page using single or small number of 'frames' so user doesn't have to learn how to read the data more than once.

- avoid the risk of just TALK TALK TALK, people learn in different ways

- Using HIGH RESOLUTION DESIGN different people can be doing different things with the same piece of data ex. the patient monitoring system, something for each specialist without concealing the other pathologies

- Avoid getting into the WAR STORY/ COUNTER STORY mode -- it is intellecutally mere up-man-ship and was part of the reason for the Challenger Disastor, as illustrated.

ISSUES: QUESTIONS TO ASK OR THINGS TO DO

1. Show Causality ~ put data in substantive order

2. Show Me The Full Data Matrix ~ if the data is pre-selected then you know you are not seeing everything. The obvious question is what's being hidden?

3. When somebody makes an argument ask: What would I really like to see if I am to believe this story? ~ pool of 1000 possible charts

Some Rules for Presentations

1) SHOW UP EARLY

2) EARLY ON SAY: Problem | Who Cares? Relevance | Solution

That's How to Begin. ex. Stumblebum technique, story about making an intentional mistake to get the attention of an audience, not recommended. Avoid distracting gestures and Never Apologize DON'T UNDERMINE YOUR OWN CREDIBILITY Try to avoid too much use of the 1st person singular. Be collegial, include your audience as peers.

3) PGP PARTICULAR | GENERAL | PARTICULAR use this rule in illustration

4) LAW: GIVE EVERY PERSON IN THE AUDIENCE AT LEAST ONE PIECE OF PAPER -- Talk Talk Talk is low bandwidth, paper is high bandwidth

5) You can know your audience by what they read! YOUR AUDIENCE IS SMART

6) You can do better than OVERHEADS (very low bandwidth)

- bullet list tends to encourage sloppy generic thinking

- ASK How is the bullet list going to be achieved?

- ex. Mission Statement generator at the Dilbert site

7) AUDIENCES ARE DESERVING OF ENDLESS RESPECT

- Assume people are your colleagues

- don't be patronizing or condescending

- they care just as much and they are just as smart

8) Use humor only in a 'pointed' fashion -- don't use longwinded STOCK JOKES and don't alienate the audience with tasteless material

9) DON'T USE MALE PRONOUNS AS UNIVERSALS instead go into the plural. The point is that there is no sense in losing half your audience.

10) BE CAREFUL ANSWERING QUESTIONS respect the questioner, if a shy audience always count to 10, plant questions to break the ice and get going

11) If you believe the stuff, let your audience know it — BE UP FRONT

12) FINISH EARLY

13) PRACTICE PRACTICE PRACTICE get help from a friend, video tape yourself, remember that rehearsals are often awful -- the big four are Quality|Relevance|Integrity|Substance

14) DRINK LOTS OF WATER the two most debilitation things are public presentations and air travel and the combination is worst of all.

15) All the other suggestions are useless if you have not stressed CONTENT. Better CONTENT is always the best suggestion.