CIS 4930/6930-002

DATA VISUALIZATION
Visual Design
Paul Rosen
Assistant Professor
University of South Florida

slides credits Miriah Meyer (U of Utah), Hanspeter Pfister (Harvard), John Stasko (Georgia Tech), & Josh Levine (U of Arizona)
 • TODAY ...
•Four Levels of Visualization Design
• Tufte’s Principles (Integrity & Design)
Critiques
•
• THE FOUR LEVELS OF
VISUALIZATION
DESIGN
 • MUNZNER’S NESTED MODEL
design model—describes levels of design
•

inherent to, and that should be considered in,

the creation of a visualization

Domain Situation
Data & Task Abstraction
Visual Encoding & Interaction
Technique & Algorithm
 • MUNZNER’S NESTED MODEL
domain situation—describing a group of
•

target users, their domain of interest, their

questions, and their data

Domain Situation
Data & Task Abstraction
Visual Encoding & Interaction
Technique & Algorithm
 • MUNZNER’S NESTED MODEL
• data/task abstraction—abstracting the specific
domain questions and data from the domain-
specific form into a generic, computational
form

Domain Situation
Data & Task Abstraction
Visual Encoding & Interaction
Technique & Algorithm
 • MUNZNER’S NESTED MODEL
• visual encoding & interaction—decide on the
specific way to create and manipulate the
visual representation of the abstraction

Domain Situation
Data & Task Abstraction
Visual Encoding & Interaction
Technique & Algorithm
 • MUNZNER’S NESTED MODEL
• algorithm—crafting a detailed procedure that
allows a computer to automatically and
efficiently carry out the desired visualization
goal

Domain Situation
Data & Task Abstraction
Visual Encoding & Interaction
Technique & Algorithm
• MUNZNER’S NESTED MODEL
 • TUFTE
• design excellence
 • TUFTE’S LESSONS
• practice—graphical integrity and excellence
• theory—design principles for data graphics
 • GRAPHICAL INTEGRITY
• clear, detailed, and thorough labeling
should be used to defeat graphical
distortion and ambiguity
 • MISSING SCALES

$521,943
$397,747
$351,341
($11,410)
$3,594,385

•TUFTE 2001
 • MISSING SCALES

$521,943
$397,747
$351,341
($11,410)
$3,594,385

baseline?

•TUFTE 2001
 • MISSING SCALES

$521,943
$397,747
$351,341
($11,410)
$3,594,385

baseline?

-$4,200,000
•TUFTE 2001
 • SCALE DISTORTION
volume

year
 • SCALE DISTORTION
volume

Did the stock market crash?

year
 • SCALE DISTORTION

scale?
volume

Did the stock market crash?

year
 • SCALE DISTORTION
volume

Did the stock market crash?

year
 • SCALE DISTORTION

context?
volume

Did the stock market crash?

year
 • SCALE DISTORTION
volume

Did the stock market crash?

year
 • SCALE DISTORTION
volume

Did the stock market crash?

year
 • TUFTE’S INTEGRITY PRINCIPLES
the representation of numbers, as physically
•

measured on the surface of the graphic itself,

should be directly proportional to the
numerical quantities represented.

size of effect shown in graphic

The Lie Factor =
size of effect in data
 • DISTORTION

!.#
Lie factor for the percent $.% +.+,
increase from 1978 to 1985 &'.! = = 5.8
-..,
()
•TUFTE 2001
•TUFTE 2001
 • TUFTE’S INTEGRITY PRINCIPLES
• show data variation, not design variation
• UNINTENDED SIZE CODING

•HTTP://PELTIERTECH.COM/WORDPRESS/BAD-BAR-CHART-PRACTICES-OR-
SEND-IN-THE-CLOWNS/
• UNINTENDED SIZE CODING

•HTTP://PELTIERTECH.COM/WORDPRESS/BAD-BAR-CHART-PRACTICES-OR-SEND-IN-THE-CLOWNS/
 • TUFTE’S INTEGRITY PRINCIPLES
• GRAPHICAL EXCELLENCE IS THAT WHICH
• gives the viewer the greatest number of ideas
in the shortest time
•

with the least ink

•

in the smallest space

•

• A. Einstein, “An explanation should be as simple as

possible, but no simpler.”
 • DESIGN PRINCIPLES
• (or how to achieve integrity and excellence)
maximize the
data-ink
Data-ink Ratio =
total ink used in graphic

TRIGLYCERIDE LEVEL
maximize the
data-ink
Data-ink Ratio =
total ink used in graphic

TRIGLYCERIDE LEVEL
maximize the
data-ink
Data-ink Ratio =
total ink used in graphic

•TUFTE 2001
COUNTER- POINT
 •EXPERIMENT
•asked participants to choose box plot with
largest range from a set
•varied representation
•measured cognitive load from EEG brain
waves
 •EXPERIMENTAL RESULTS
•studies showed that the simplest (highest
data-ink ratio) box plot is hardest to
interpret
•paper focused on cognitive load as an
evaluation method
• AVOID CHART JUNK
• AVOID CHART JUNK
• AVOID CHART JUNK
• AVOID CHART JUNK
• AVOID CHART JUNK
• ATTRACTION OR DISTRACTION?

•NIGEL HOLMES,TIME MAGAZINE

• COUNTER-POINTS
 • EXPERIMENTAL QUESTIONS
do visual embellishments cause
•

comprehension problems?
do embellishments provide additional
•

information that is valuable for the reader?

 • EXPERIMENTAL RESULTS
• No significant difference between plain and
embellished charts for interactive
interpretation accuracy
No significant difference in recall accuracy
•

after a five-minute gap

56
 • EXPERIMENTAL RESULTS
Significantly better recall for embellished charts of
•

both the chart topic and the details (categories and

trend) after long-term gap (2-3 weeks)
Participants saw value messages in the embellished
•

charts significantly more often than in the plain charts

Participants found the embellished charts more
•

attractive, most enjoyed them, and found that they

were easiest and fastest to remember

56
 What Makes a Visualization Memorable?
Michelle A. Borkin, Student Member, IEEE, Azalea A. Vo, Zoya Bylinskii, Phillip Isola, Student Member, IEEE,
Shashank Sunkavalli, Aude Oliva, and Hanspeter Pfister, Senior Member, IEEE

Fig. 1. Left: The top twelve overall most memorable visualizations from our experiment (most to least memorable from top left to
bottom right). Middle: The top twelve most memorable visualizations from our experiment when visualizations containing human
recognizable cartoons or images are removed (most to least memorable from top left to bottom right). Right: The twelve least
memorable visualizations from our experiment (most to least memorable from top left to bottom right).

Abstract—An ongoing debate in the Visualization community concerns the role that visualization types play in data understanding.
In human cognition, understanding and memorability are intertwined. As a first step towards being able to ask questions about impact
and effectiveness, here we ask: “What makes a visualization memorable?” We ran the largest scale visualization study to date using
2,070 single-panel visualizations, categorized with visualization type (e.g., bar chart, line graph, etc.), collected from news media sites,
government reports, scientific journals, and infographic sources. Each visualization was annotated with additional attributes, including
ratings for data-ink ratios and visual densities. Using Amazon’s Mechanical Turk, we collected memorability scores for hundreds of
these visualizations, and discovered that observers are consistent in which visualizations they find memorable and forgettable. We
find intuitive results (e.g., attributes like color and the inclusion of a human recognizable object enhance memorability) and less
intuitive results (e.g., common graphs are less memorable than unique visualization types). Altogether our findings suggest that
quantifying memorability is a general metric of the utility of information, an essential step towards determining how to design effective
visualizations.
Index Terms—Visualization taxonomy, information visualization, memorability
 • RESULTS
color and human recognizable objects
•

enhance memorability
• common graphs are less memorable than
unique visualization types
•CHART JUNK? IT DEPENDS
•persuasion
•memorability PROS
•engagement

•unbiased analysis
•trustworthiness
CONS
•interpretability
•space efficiency

59
•TUFTE 2001
maximize the
number of entries in data array
Data Density =
area of data graphic
 • SHRINK THE GRAPHICS
with small multiples
•

g1 m2 g4 m3 g5 g6 g7 m4 g13 g12 m7 g16 g17 g18 g20 g19 g18 g22 g23 g24 g25 g28 g31

•MEYER ET AL 2010
• SHRINK THE GRAPHICS

•BERTIN 1967
• SHRINK THE GRAPHICS
•with sparklines

•TUFTE 2006
• COUNTER-POINT
 • ILLUSIONS OF VISUAL
BANDWIDTH
• people over-predict what they will see and
become aware of
 • OVERESTIMATE OF BREADTH
• belief that viewers can take in all (or most) of
the details of a scene at once
•adding extra visual features makes it harder
to find specifics bits of information

74
 • OVERESTIMATE OF COUNTENANCE
belief that user will attend to a higher
•

proportion of the display than they do

• users typically have expectations about where
in a display to look

74
 • OVERESTIMATE OF DEPTH
belief that attending to an object leads to
•

more complete and deep understanding than

is the case

74
 • TUFTE’S DESIGN PRINCIPLES
maximize the data-ink ratio
•

avoid chart junk (sometimes)

•

use multifunctioning elements

•

layer information
•

maximize the data density

•

shrink the graphics

•

• maximize the amount of data shown

(sometimes)
• DESIGN CRITIQUES
•HTTP://WWW.GUARDIAN.CO.UK/TECHHNOLOGY/BLOG/2008/JAN/21/LIESDAMNLIESANDSTEVEJOBS
• U.S. SMARTPHONE MARKETSHARE

RIM
Apple
Palm
Motorola
Nokia
Other
• U.S. SMARTPHONE MARKETSHARE

21.2%

39.0%
RIM
Apple
3.1%
Palm
7.4% Motorola
Nokia
Other
9.8%

19.5%
 • U.S. SMARTPHONE MARKETSHARE
40%
39.0

30%

20% 21.2
19.5

10%
9.8
7.4

3.1
0%
RIM Other Apple Palm Motorola Nokia
 • RECOMMENDED READING
• Visualization Analysis & Design: Chapter 4 (pp. 66-93)
•The Visual Display of Quantitative Information: all