From Hierarchies to Polyarchies Visualizing Multiple Relationships George

From Hierarchies to Polyarchies: Visualizing Multiple Relationships George G. Robertson Microsoft Research

What is the problem? t Hierarchies are very common t 20 years of hierarchy visualization R&D t Significant problems remain t New problems appearing (Multiple Hierarchies) GVU 11/2000 2

Current Approaches Many 2 D and 3 D hierarchy visualizations t Each works for some tasks and some scales t Very few have had user testing Windows Tree Control t Many observed problems GVU 11/2000 3

What’s wrong with this picture? GVU 11/2000 4

Problems: Fitting Data t Extreme aspect ratio (broad and shallow) t May be multiple hierarchies t Scaling issues GVU 11/2000 5

Problems: Cognitive Overhead t Loss of context ²Or loss of detail ²Separate detail/overview extra overhead t Multiple focus is difficult t Which item is open? GVU 11/2000 6

Basic View Strategies Two view (separate detail/overviews) Distorted view t Distorted data: fisheye t Distorted space: 3 D, hyperbolic Focus in Context (integrated view) GVU 11/2000 7

Basic Visualization Approaches Indentation t Tree control t Fisheye Clustering t Galaxy of News t Theme. Scape Containment t Hot Sauce t Treemaps t Pad++ Node-link diagrams Geographic t Floor plans t 2 D diagrams t Sem. Net t Cone Tree t Fisheye Cone Tree t Hyperbolic viewer t FSN t XML 3 D t Street maps GVU 11/2000 8

Smalltalk File Browser - 1979 t Xerox PARC 1979 GVU 11/2000 9

Fisheye Views – 1986 t Furnas, Bell Labs report 1981, CHI’ 86 GVU 11/2000 10

Sem. Net - 1986 t Fairchild, Poltrock, & Furnas, MCC t Semantic network 3 D visualization GVU 11/2000 11

Cone Tree - 1991 t Robertson, Mackinlay & Card, Xerox PARC, CHI’ 91 t Limits: ² 10 levels ² 1000 nodes ²Up to 10, 000 GVU 11/2000 12

Cone Tree – 10, 000 nodes t Performance problems and Visual clutter GVU 11/2000 13

Fisheye Cone Tree t Adding Fisheye solves problems GVU 11/2000 14

Tree. Map - 1991 t Johnson & Shneiderman, U. Maryland, Vis’ 91 t Space filling t ~3000 objects t Micro. Logic’s Disk. Mapper GVU 11/2000 15

FSN - 1992 t Tesler, SGI, 1992 t 2 D layout on 3 D surface t Loss of context when examining detail t Hyper-G (Graz; Andrews; Info. Vis’ 95) GVU 11/2000 16

Graphical Fisheye - 1992 t Sarkar & Brown, DEC SRC, CHI’ 92 GVU 11/2000 17

Rubber Sheet - 1993 t Sarkar & Snibbe, Brown, UIST’ 93 GVU 11/2000 18

Spiral Visualization - 1994 t Mackinlay, Robertson, & De. Line, Xerox PARC, UIST’ 94 t Large DAGs t Only shows current path GVU 11/2000 19

Hyperbolic Browser - 1994 t Lamping & Rao, Xerox PARC, UIST’ 94 t Projected onto circle t 1000’s of nodes GVU 11/2000 20

Continuous Zoom - 1995 t Bartram et al, Simon Fraser, UIST’ 95) GVU 11/2000 21

Fsviz - 1995 t Carriere and Kazman, Waterloo, Info. Vis’ 95 t Up to 5000 nodes with no occlusion GVU 11/2000 22

Butterfly - 1995 t Mackinlay, Xerox PARC, CHI’ 95 GVU 11/2000 23

Index View - 1995 t T. Masui, Sharp, UIST 95 t Vertical picks focus t Horizontal controls zoom Aaaamon. dll Ksproxy. ax Ksqmf. ax Edb 500. dll Ksuser. dll Ksproxy. ax Ksuser. dll Ksvpintf. ax Label. exe Lmrt. dll Label. exe Legacy. inf Lmrt. dll Lanman. drv Oakley. dll Legacy. inf Linkinfo. dill Tapi. dll Lmrt. dll Wavemsp. dll GVU 11/2000 24

H 3 - 1997 t Munzner, Stanford Univ. , Info. Vis’ 97 t Projected onto sphere t 20, 000 nodes GVU 11/2000 25

Reconfigurable Disk Tree - 1998 t Jeong and Pang, UC Santa Cruz, Info. Vis’ 98 t Number of identifiable nodes: 3000 – 4000 GVU 11/2000 26

Disk Tree - 1998 t Chi et al, Xerox PARC, CHI’ 98 t Compact 2 D representation GVU 11/2000 27

Sunburst - 2000 t Stasko & Zhang, Georgia Tech, Info. Vis 2000 t Radial space-filling t Techniques for viewing more detail GVU 11/2000 28

Visualization Taxonomy - 1994 Noik (Graphics Interface’ 94) t Implicit (use of perspective) ²Continuous focus and context t Filtered (removing items of low interest) ²Discrete focus and context t Distorted (size, shape, position of elements) t Adorned (color, texture) GVU 11/2000 29

Unresolved Problems t Multiple focus GVU 11/2000 30

Folding Paper - proposal GVU 11/2000 31

Multiple Focus in 3 D - proposal GVU 11/2000 32

Info. Bowl - proposal GVU 11/2000 33

Unresolved Problems t Multiple focus t Multiple hierarchies GVU 11/2000 34

Multiple Hierarchies t One hierarchy changing over time ²Time Tube ²Taxonomy visualization t Multi. Trees (shared subtrees) ²XML 3 D t Polyarchy (multiple intersecting hierarchies) GVU 11/2000 35

Time Tube - 1998 t Chi et al. , Xerox PARC, CHI’ 98 GVU 11/2000 36

Taxonomy Visualization - 2000 t Graham et al. , Napier Univ. IJHCS 2000 GVU 11/2000 37

Multi. Trees - 1994 t Furnas & Zacks, Bell Core, CHI’ 94 t DAG’s with shared subtrees GVU 11/2000 38

XML 3 D - 2000 t Munzner, Stanford, IJHCS 2000 GVU 11/2000 39

Polyarchy Visualization Problem: People and Resources Example Multiple Hierarchies Exist t Direct reporting t Cost or Profit Center t Location But only one t Implicit relationships hierarchy is shown GVU 11/2000 40

Show Other Hierarchies Find: Daniel Robbins t Naming hierarchy is default t Allow maintainer to define hierarchies t Show search results in current view t Morph between views GVU 11/2000 41

Visualization Opportunities t Perceptual cues to provide pattern information t User can search or browse t Show hierarchies an object participates in t Show relationships between hierarchies t Show relationships between objects GVU 11/2000 42

One Selection, One Hierarchy GVU 11/2000 43

Visual Pivot – Other Hierarchies GVU 11/2000 44

Relationships Between People GVU 11/2000 45

Relationships Between Multiple People GVU 11/2000 46

Unresolved Problems t Multiple focus t Multiple hierarchies t Evaluate hierarchy visualizations GVU 11/2000 47

Polyarchy Visualization User Studies t Study 1: Mockup of visual pivot ²Issues list guided development of prototype t Study 2: Prototype: 2 D vs 3 D ²Visual Pivot animation was misleading ²Animation Speeds were too slow GVU 11/2000 48

Polyarchy Visualization User Studies t Study 3: Animation Styles and Speeds ²Six animation styles: Picked two best ²Twice as fast as study 2: Still too slow t Study 4: Prototype: 2 D vs 3 D ²Identified most effective animation style ²Identified best speed range GVU 11/2000 49

Two Styles of Visual Pivot t Sliding t Rotating GVU 11/2000 50

Visual Pivot Styles Sliding versus Rotating GVU 11/2000 51

Visual Pivot Styles Learning Effects GVU 11/2000 52

Visual Pivot Speed GVU 11/2000 53

Unresolved Problems t Multiple focus t Multiple hierarchies t Evaluate hierarchy visualizations t Get the solutions out to real users! GVU 11/2000 54

Summary Use perception to reduce cognitive overhead Need continued research to solve problems t Help user focus on multiple items of interest without losing context t Multiple hierarchies and other relationships t Handle larger scale Evaluate Iterate get results out GVU 11/2000 55