why does this suck?

why does this suck?

Information Visualization

Jeffrey HeerUC Berkeley | PARC, Inc.

CS160 – 2004.11.22

(includes numerous slides from Marti Hearst, Ed Chi, Stuart Card, and Peter Pirolli)

Basic Problem

We live in a new ecology.

Scientific JournalsJournals/person increases 10X every 50 yearsJournals/person increases 10X every 50 years

YearYear

0.01

0.1

1

10

100

1000

10000

100000

1000000

1750 1800 1850 1900 1950 2000

JournalsJournals

Journals/People x10Journals/People x1066

DarwinDarwin V. BushV. Bush YouYou

Web Ecologies

1

10

100

1000

10000

100000

1000000

10000000

Aug-92 Feb-93 Aug-93 Feb-94 Aug-94 Feb-95 Aug-95 Feb-96 Aug-96 Feb-97 Aug-97 Feb-98 Aug-98

Serv

ers

Source: World Wide Web Consortium, Mark Gray, Netcraft Server Survey

1 new server every 2 seconds7.5 new pages per second

Human Capacity

0.01

0.1

1

10

100

1000

10000

100000

1000000

1750 1800 1850 1900 1950 2000DarwinDarwin V. BushV. Bush YouYou

Attentional Processes“What information consumes is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention, and a need to allocate that attention efficiently among the overabundance of information sources that might consume it.”

~Herb Simonas quoted by Hal Varian

Scientific AmericanSeptember 1995

Human-Information Interaction The real design problem is not increased

access to information, but greater efficiency in finding useful information.

Increasing the rate at which people can find and use relevant information improves human intelligence.

Amount ofAccessibleKnowledge

Amount ofAmount ofAccessibleAccessibleKnowledgeKnowledge

Cost [Time]Cost [Time]Cost [Time]

Amount ofAccessibleKnowledge

Amount ofAmount ofAccessibleAccessibleKnowledgeKnowledge

Cost [Time]Cost [Time]Cost [Time]

Information Visualization Leverage highly-developed human visual

system to achieve rapid understanding of abstract information.

1.2 b/s (Reading)2.3 b/s (Pictures)

Information Visualization “Transformation of the symbolic into the geometric”

(McCormick et al., 1987)

“... finding the artificial memory that best supports our natural means of perception.'‘ (Bertin, 1983)

The depiction of information using spatial or graphical representations, to facilitate comparison, pattern recognition, change detection, and other cognitive skills by making use of the visual system. (Hearst, 2003)

Why Visualization? Use the eye for pattern recognition; people good at

scanning recognizing remembering images

Graphical elements facilitate comparisons via length shape orientation texture

Animation shows changes across time Color helps make distinctions Aesthetics make the process appealing

Visualization Success Stories

Visualization Success Story

Mystery: what is causing a cholera epidemic in London in 1854?

Visualization Success Story

From Visual Explanations by Edward Tufte, Graphics Press, 1997

Illustration of John Snow’sdeduction that a cholera epidemic was caused by a bad water pump, circa 1854.

Horizontal lines indicate location of deaths.

Visualization Success Story

From Visual Explanations by Edward Tufte, Graphics Press, 1997

Illustration of John Snow’s deduction that a cholera epidemic was caused by a bad water pump, circa 1854.

Horizontal lines indicate location of deaths.

A Visualization Expedition

(a tour through past and present)

Perspective Wall

Slide adapted from Chris North 18

Starfield Displays

Film Finder

Table Lens

Distortion Techniques

Indented Hierarchy Layout

Places all items along vertically spaced rows

Uses indentation to show parent child relationships

Breadth and depth end up fighting for space resources

Top-down layout

Uses separate dimensions for breadth and depth

Reingold-Tilford Layout

tidier drawing of trees - reingold, tilford

TreeMaps

Space-filling technique that divides space recursively

Segments space according to ‘size’ of children nodes

map of the market – smartmoney.com

SpaceTree

Cone Trees

Tree layout in three dimensions

Shadows provide 2D structure

Can also make “Balloon Trees” – 2D version of ConeTree

cone tree – robertson, mackinlay, and card

Degree-of-Interest Trees

Hyperbolic Trees

Network visualization

Often uses physics models (e.g., edges as springs) to perform layout.

Can be animated and interacted with.

Network Visualization

Skitter, www.caida.org

WebBook

Web Forager

Document Lens

Data Mountain

Supports document organization in a 2.5 dimensional environment.

Designing Visualizations

(some tricks of the trade)

Graphical Excellence [Tufte]

the well-designed presentation of interesting data – a matter of substance, of statistics, and of design

consists of complex ideas communicated with clarity, precision and efficiency

is that which gives to the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space

requires telling the truth about the data.

Interactive Tasks [Shneiderman]

1. Overview: Get an overview of the collection2. Zoom: Zoom in on items of interest3. Filter: Remove uninteresting items4. Details on demand: Select items and get

details5. Relate: View relationships between items6. History: Keep a history of actions for undo,

replay, refinement7. Extract: Make subcollections

Proposed Data Types1. 1D: timelines,…2. 2D: maps,…3. 3D: volumes,…4. Multi-dimensional: databases,…5. Hierarchies/Trees: directories,…6. Networks/Graphs: web,…7. Document collections: digital libraries,…

This is useful, but what’s wrong here?

Basic Types of Data Nominal (qualitative)

(no inherent order) city names, types of diseases, ...

Ordinal (qualitative) (ordered, but not at measurable intervals) first, second, third, … cold, warm, hot Mon, Tue, Wed, Thu …

Interval (quantitative) integers or reals

QUANT ORDINAL NOMINAL

Position Position PositionLength Density Color HueAngle Color Saturation TextureSlope Color Hue ConnectionArea Texture ContainmentVolume Connection DensityDensity Containment Color SaturationColor Saturation Length ShapeColor Hue Angle Length

Ranking of Applicability of Properties for Different Data Types(Mackinlay 88, Not Empirically Verified)

Visualization Design Patterns Pre-Attentive Patterns

Leverage things that automatically “pop-out” to human attention Stark contrast in color, shape, size, orientation

Gestalt Properties Use psychological theories of visual grouping proximity, similarity, continuity, connectedness, closure,

symmetry, common fate, figure/ground separation High Data Density

Maximize number of items/area of graphic This is controversial! Whitespace may contribute to good visual

design… so balance appropriately. Small Multiples

Show varying visualizations/patterns adjacent to one another Enable Comparisons

Visualization Design Patterns Focus+Context

Highlight regions of current interest, while de-emphasizing but keeping visible surrounding context.

Can visually distort space, or use degree-of-interest function to control what is and isn’t visualized.

Dynamic Queries Allow rapid refinement of visualization criteria Range sliders, Query sliders

Panning and Zooming Navigate large spaces using a camera metaphor

Semantic Zooming Change content presentation based on zooming level Hide/reveal additional data in accordance with available space

Software Architectures The Information Visualization Reference

Model [Chi, Card, Mackinlay, Shneiderman]

Evaluating Visualizations

Evaluating Visualizations Visualizations are user interfaces, too…established

methodologies can be used. Questions to ask

What tasks do you expect people to perform with the visualization?

What interfaces currently exist for this task? In what ways do you expect different visualizations to help

or hurt aspects of these tasks? Metrics: task time, success rate, information gained

(e.g., test the user, or exploit priming effects), eye tracking.

Evaluating Hyperbolic Trees The Great CHI’97 Browse-Off: Individual

browsers race against the clock to perform various retrieval and comparison tasks.

Hyperbolic Tree won against M$ File Explorer and others.

Can we conclude that it is the better browser?

vs.

Evaluating Hyperbolic Trees No!

Different people operating each browser. Tasks were not ecologically valid.

Can’t say what is better for what. PARC researchers did extensive eye-tracking studies uncovering

very nuanced visual psychology. Found Hyperbolic Tree is better when underlying information

design (e.g., tree structure and labeling) is better. In case of CHI Browse Off, the Hyperbolic Tree had a quicker

human user “behind the wheel”. Moral: Exercise judicious study design, but also don’t feel let

down if task times are not being radically improved… subtleties abound.

Questions?

Jeffrey Heer jheer@cs.berkeley.eduprefuse http://prefuse.sourceforge.net

Accuracy Ranking of Quantitative Perceptual TasksEstimated; only pairwise comparisons have been validated

(Mackinlay 88 from Cleveland & McGill)

Interpretations of Visual PropertiesSome properties can be discriminated more accurately but don’t have intrinsic meaning

Density (Greyscale)Darker -> More

Size / Length / AreaLarger -> More

PositionLeftmost -> first, Topmost -> first

Hue??? no intrinsic meaning

Slope??? no intrinsic meaning

Micro-Aspects of Visualization Design(aka fun with visual psychology)

Preattentive Processing A limited set of visual properties are

processed preattentively (without need for focusing attention).

This is important for design of visualizationswhat can be perceived immediatelywhat properties are good discriminatorswhat can mislead viewers

All Preattentive Processing figures from Healey 97http://www.csc.ncsu.edu/faculty/healey/PP/PP.html

Example: Color Selection

Viewer can rapidly and accurately determinewhether the target (red circle) is present or absent.Difference detected in color.

Example: Shape Selection

Viewer can rapidly and accurately determinewhether the target (red circle) is present or absent.Difference detected in form (curvature)

Pre-attentive Processing < 200 - 250ms qualifies as pre-attentive

eye movements take at least 200ms yet certain processing can be done very quickly,

implying low-level processing in parallel If a decision takes a fixed amount of time

regardless of the number of distractors, it is considered to be preattentive.

Example: Conjunction of Features

Viewer cannot rapidly and accurately determinewhether the target (red circle) is present or absent when target has two or more features, each of which arepresent in the distractors. Viewer must search sequentially.

All Preattentive Processing figures from Healey 97http://www.csc.ncsu.edu/faculty/healey/PP/PP.html

Example: Emergent Features

Target has a unique feature with respect to distractors (open sides) and so the groupcan be detected preattentively.

Example: Emergent Features

Target does not have a unique feature with respect to distractors and so the group cannot be detected preattentively.

Asymmetric and Graded Preattentive Properties Some properties are asymmetric

a sloped line among vertical lines is preattentive a vertical line among sloped ones is not

Some properties have a gradation some more easily discriminated among than others

Use Grouping of Well-Chosen Shapes for Displaying Multivariate Data

SUBJECT PUNCHED QUICKLY OXIDIZED TCEJBUS DEHCNUP YLKCIUQ DEZIDIXOCERTAIN QUICKLY PUNCHED METHODS NIATREC YLKCIUQ DEHCNUP SDOHTEMSCIENCE ENGLISH RECORDS COLUMNS ECNEICS HSILGNE SDROCER SNMULOCGOVERNS PRECISE EXAMPLE MERCURY SNREVOG ESICERP ELPMAXE YRUCREMCERTAIN QUICKLY PUNCHED METHODS NIATREC YLKCIUQ DEHCNUP SDOHTEMGOVERNS PRECISE EXAMPLE MERCURY SNREVOG ESICERP ELPMAXE YRUCREMSCIENCE ENGLISH RECORDS COLUMNS ECNEICS HSILGNE SDROCER SNMULOCSUBJECT PUNCHED QUICKLY OXIDIZED TCEJBUS DEHCNUP YLKCIUQ DEZIDIXOCERTAIN QUICKLY PUNCHED METHODS NIATREC YLKCIUQ DEHCNUP SDOHTEMSCIENCE ENGLISH RECORDS COLUMNS ECNEICS HSILGNE SDROCER SNMULOC

SUBJECT PUNCHED QUICKLY OXIDIZED TCEJBUS DEHCNUP YLKCIUQ DEZIDIXOCERTAIN QUICKLY PUNCHED METHODS NIATREC YLKCIUQ DEHCNUP SDOHTEMSCIENCE ENGLISH RECORDS COLUMNS ECNEICS HSILGNE SDROCER SNMULOCGOVERNS PRECISE EXAMPLE MERCURY SNREVOG ESICERP ELPMAXE YRUCREMCERTAIN QUICKLY PUNCHED METHODS NIATREC YLKCIUQ DEHCNUP SDOHTEMGOVERNS PRECISE EXAMPLE MERCURY SNREVOG ESICERP ELPMAXE YRUCREMSCIENCE ENGLISH RECORDS COLUMNS ECNEICS HSILGNE SDROCER SNMULOCSUBJECT PUNCHED QUICKLY OXIDIZED TCEJBUS DEHCNUP YLKCIUQ DEZIDIXOCERTAIN QUICKLY PUNCHED METHODS NIATREC YLKCIUQ DEHCNUP SDOHTEMSCIENCE ENGLISH RECORDS COLUMNS ECNEICS HSILGNE SDROCER SNMULOC

Text NOT PreattentiveText NOT Preattentive

Preattentive Visual Properties(Healey 97)

length Triesman & Gormican [1988] width Julesz [1985] size Triesman & Gelade [1980] curvature Triesman & Gormican [1988] number Julesz [1985]; Trick & Pylyshyn [1994] terminators Julesz & Bergen [1983] intersection Julesz & Bergen [1983] closure Enns [1986]; Triesman & Souther [1985] colour (hue) Nagy & Sanchez [1990, 1992]; D'Zmura [1991]

Kawai et al. [1995]; Bauer et al. [1996] intensity Beck et al. [1983]; Triesman & Gormican [1988] flicker Julesz [1971] direction of motion Nakayama & Silverman [1986]; Driver & McLeod [1992] binocular lustre Wolfe & Franzel [1988] stereoscopic depth Nakayama & Silverman [1986] 3-D depth cues Enns [1990] lighting direction Enns [1990]

Gestalt Principles

Idea: forms or patterns transcend the stimuli used to create them. Why do patterns emerge? Under what circumstances?

Principles of Pattern Recognition “gestalt” German for “pattern” or “form, configuration” Original proposed mechanisms turned out to be wrong Rules themselves are still useful

Gestalt PropertiesProximity

Why perceive pairs vs. triplets?

Gestalt PropertiesSimilarity

Slide adapted from Tamara Munzner

Gestalt PropertiesContinuity

Slide adapted from Tamara Munzner

Gestalt PropertiesConnectedness

Slide adapted from Tamara Munzner

Gestalt PropertiesClosure

Slide adapted from Tamara Munzner

Gestalt PropertiesSymmetry

Slide adapted from Tamara Munzner

Gestalt Laws of Perceptual Organization (Kaufman 74)

Figure and Ground Escher illustrations are good examples Vase/Face contrast

Subjective Contour

More Gestalt Laws Law of Common Fate

like preattentive motion propertymove a subset of objects among similar ones and

they will be perceived as a group

Colors for Labeling Ware recommends to take into account:

Distinctness Unique hues

Component process model Contrast with background Color blindness Number

Only a small number of codes can be rapidly perceived Field Size

Small changes in color are difficult to perceive Conventions

Ware’s Recommended Colors for Labeling

Red, Green, Yellow, Blue, Black, White, Pink, Cyan, Gray, Orange, Brown, Purple.The top six colors are chosen because they are the unique colors that mark the ends of the opponent color axes. The entire set corresponds to the eleven color names found to be the most common in a cross-cultural study, plus cyan (Berlin and Kay)