Some Visualization Theory

Visual Analytics for Linguists

ESSLLI 2014

Chris Culy

revised 7 August 2014

3 Possible goals for a visualization

Exploration
Analysis ("Comprehension")
Presentation ("Communication")

Pipeline Model

The (conceptual) structure of a visualization

Reference:
S. Card, J. Mackinlay, and B. Shneiderman. 1999. Readings in Infomration Visualization: Using Vision to Think. Chapter 1.

3 Key ideas for visual representations

Gestalt Principles:
The tendency to perceive elements as belonging to a group, based on certain visual properties
Pre-attentiveness:
Certain low level visual aspects are recognized before conscious awareness
Visual variables:
The different visual aspects that can be used to encode information

Gestalt Proximity/Density

"Things that are close together are perceptually grouped together."

•	•	•
•	•	•
•	•	•

•	•	•
•	•	•
•	•	•

Elements of similar density are perceptually grouped together

References:
C. Ware. 2012. Information Visualization: Perception for Design. Chapter 6.

Gestalt Similarity

Things with similar shape, texture, or color tend to be grouped together

• • • • •
X X O X X O X

References:
C. Ware. 2012. Information Visualization: Perception for Design. Chapter 6.

Gestalt Relative Size

Things with similar size tend to be grouped together

• • • • •

References:
C. Ware. 2012. Information Visualization: Perception for Design. Chapter 6.

Gestalt Connectedness

Things connected by lines are (strongly) grouped.

References:
C. Ware. 2012. Information Visualization: Perception for Design. Chapter 6.

Gestalt Symmetry

"Symmetry can provide a strong organizing principle."
Notes: "most sensitive" to small, vertically oriented
Especially reflectional symmetry and violations of it

"Hot coffee" in American (left) and British (right) English

References:
C. Ware. 2012. Information Visualization: Perception for Design. Chapter 6.
Graphs from Google Ngram Viewer

Other Gestalt Principles

Continuity: "we are more likely to construct visual entities out of visual elements that are smooth and continuous, rather than ones that contain abrupt changes in direction"
Consequence: curved lines are easier to follow than bent lines
Closure: "A closed contour tends to be seen as an object" (and we tend to close contours with gaps)
Common fate: Things with similar motion tend to be grouped (within limits)

References:
C. Ware. 2012. Information Visualization: Perception for Design. Chapter 6.

XXXXXXXXXXXX
XXXXXXXXXXXX
XXXXXXXXXOXX
XXXXXXXXXXXX

Pre-attentive, aka "pop-out", features

Certain visual aspects are very prominent, especially when they are relatively isolated against a background

"Low level" visual features are processed sooner and faster than higher level patterns
in V1 part of the cortex
color, grey, shape, size, length, orientation, motion, additional mark (+ others)
Recognition is facilitated (biased) by higher level task and attention

References:
C. Ware. 2010. Visual Thinking for Design. Chapter 2

How to use preattentive features

Target item(s) differ from the other items in the value of one (or more) low level feature
Target items are few (1 is best)
Other items and the background are relatively homogeneous
Possible to have 2 separate targets, using 2 different features, one for each target
Motion is a strong attention-getter

References:
C. Ware. 2010. Visual Thinking for Design. Chapter 2

/ / / / / / / / /
/ / / / / / / / /
/ / / \ / / / / /
/ / / / / / / / /

Look for these Os on the next slide

X X O X X O X X O
O X X X O O X X O
X X X O X O X O O
X O X O X O O X X

More details

We are better at recognizing contrasts than absolute values
Only ~ 3 levels of difference available for "pop out"
Differences are not symmetric,
e.g. bigger than others is stronger than smaller than others
Detail (and most preattentive features) are best in the center of vision
motion is detectable everywhere

References:
C. Ware. 2010. Visual Thinking for Design. Chapter 2

Practically speaking ...

Watch out for unintended Gestalt effects
Put related elements near each other, with most important in center
Make differences in visual values as large as possible
Use positive difference for highlighting

References:
C. Ware. 2012. Information Visualization: Perception for Design. Appendix D.

Types of data

Data can be best visualized according to its type

Nominal: no intrinsic order, e.g. words
Ordinal: ordered in a sequence, e.g. words ordered by orthographic representation
Interval: ordinal where gaps make sense, e.g. time
Ratio: ordinal where proportions make sense, e.g. frequencies

References:
C. Ware. 2012. Information Visualization: Perception for Design. Chapter 1.

Marks and Visual variables

Jacques Bertin:

A mark is "something that is visible and can be used ... to show relationships within sets of data"
Visual variables are "the different ways that a mark can be varied"

Reference:
S. Carpendale. 2003. Considering Visual Variables as a Basis for Visualization. Research report 2001-693-16, Department of Computer Science, University of Calgary.

Marks: realizations of concepts

Type	Concept	Realization
Point	Location only	By shape, size, color ...
Line	Location + length	By width, color ...
Area	Length + width in 2D	By color ...; also has size
Surface	Length + width in 3D	ditto
Volume	Length + width + depth	ditto

Reference: Summary of:
S. Carpendale. 2003. Considering Visual Variables as a Basis for Visualization. Research report 2001-693-16, Department of Computer Science, University of Calgary.

What are the (useful) visual variables?

Mostly the low level features, since higher level patterns take longer to process

Carpendale gives these as primary:
Position, size, shape, value (brightness), color (hue), orientation, grain (density), pattern, texture (surface appearance), motion

There are others that are less used, e.g. transparency, blur, occlusion ...

Reference: Summary of:
S. Carpendale. 2003. Considering Visual Variables as a Basis for Visualization. Research report 2001-693-16, Department of Computer Science, University of Calgary.

Characteristics of visual variables

These help us figure out how to use them

Selective. CuC: Is X different from the others?
Associative. CuC: Is X like the others?
Order. CuC: Is X more/greater/bigger/... than Y
Quantitative. CuC: How much is the difference between X and Y?
Length. CuC: How many different categories can we represent with this variable for a task?

Reference: Summary of:
S. Carpendale. 2003. Considering Visual Variables as a Basis for Visualization. Research report 2001-693-16, Department of Computer Science, University of Calgary.

Visual variables

?
?
?
?
?
?
?
?
?
?

Reference: Summary of:
S. Carpendale. 2003. Considering Visual Variables as a Basis for Visualization. Research report 2001-693-16, Department of Computer Science, University of Calgary.

Practically speaking ...

Don't use ordered variables for nominal data
Use position for the most important information
Many lengths are ~5 — a limiting factor
For quantity, use (vertical) position, length, size, lightness, saturation(?)
For many distinctions: use size, shape or brightness
Redundant encoding may make features easier to interpret
but cf. Edward Tufte

Notes about color

Color is powerful, but there are special considerations

Hue is nominal, but can be given (cultural) orderings (e.g. temperature, elevation)
Affective connotations of color vary across cultures
Depending on who your users are, you may need to design for colorblindness or low-vision
Allowing users to set colors / color schemes is the most flexible.

Resource:
Sim Daltonism: color blindness simulator for OS X
(There other tools, and for other platforms)

Practically speaking

Use lighter, less saturated colors for larger areas
Use darker, more saturated colors for smaller areas and lines
For ordinal data, use darker and more saturated for higher values (e.g. heat map)
Color (hue) length is ~6-10, so with more categories, color may not be an appropriate visual variable.
Saturation has a length of ~3
Ensure a luminance contrast of 3:1 for text.

Resource:
Color Brewer

How should visualizations interact with the user?

Shneiderman's "Information Seeking Mantra"

Overview first. Zoom and Filter. Details on Demand

Keim's Visual Analytics Mantra

Analyze first, show the important, zoom, filter and analyze further, details on demand

References:
B. Shneiderman. 1996. The eyes have it: A task by data type taxonomy for information visualizations. Proceedings of the 1996 IEEE Sympoium on Visual Languages, VL '96.
D. A. Keim, F. Mansmann, J. Schneidewind, and H. Ziegler. 2006. Challenges in visual data analysis. Proceedings of the conference on Information Visualization, IV '06.

What do users (want to) do?

Select: mark something as interesting
Explore: show me something else
Reconfigure: show me a different arrangement [same type]
Encode: show me a different representation [different type]
Abstract/Elaborate: show me less or more detail
Filter: show me something conditionally
Connect: Show me related items

Some other things: history, annotate, extract

References:
J.S. Yi, Y.A. Kange, J.T. Sasko, J.A. Jacko. 2007. Toward a deeper understanding of the role of interaction in information visualization. IEEE Transactions on Visualization and Computer Graphics. 13:6
B. Shneiderman. 1996. The eyes have it: A task by data type taxonomy for information visualizations. Proceedings of the 1996 IEEE Sympoium on Visual Languages, VL '96.