Transcript lynch_pres1

Using Vision to Think
Stuart Card, Jock Mackinlay, Ben Shneiderman
Knowledge Crystallization
Task to gain information, then
package it for communication or
action
Consider buying a laptop:
Search the internet for data
Identify what’s most interesting in the data
Create a representation to be able to see data
better
Make decision
Knowledge Crystallization(2)
Five typical items of knowledge crystallization
in scenario:
Information foraging
Search for schema(representation)
Instantiate schema with data
Problem-solve trade-offs
Search for new schema
Package patterns
Information Visualization
“The use of computer-supported, interactive,
visual representations of abstract data to
amplify cognition”
“Information visualization is a new emerging field whose goal is the
perceptualization of information. Information visualization differs from
scientific and medical visualization in that the data to be visualized is
inherently non-spatial. (Robertson)”
• “ Information visualization in enables users to get information quickly,
put it in a meaningful shape, and to make decisions in a relatively short
time.”
Origins of InfoVis
Data graphics community: Playfair, Tufte,
Bertin
Tukey’s “Exploratory Data Analysis”
Cleveland and McGill’s “Dynamic Graphics for
Statistics”
Origins(2)
NSF initiative on scientific visualization(1985)
Mackinlay’s APT
Advancements in user interface community
Early Infovis
Active Diagrams
Large-scale data
monitoring
Mapping Data to Visuals
Data Tables
2D table in which columns represent cases,
and variables as rows
Data Table Variables
There are three basic types of variables for
data tables:
N = Nominal
O = Ordinal
Q = Quantitative
Metadata
Descriptive information about data, or any
element of data
One variable that can contain more than one
piece of information
Ex: x,y,z can just be one variable, position.
Data Transformations
Transform data into a new form, higher
abstraction
Four types:
Values -> Derived Values
Structure -> Derived Structure
Values -> Derived Structure
Structure -> Derived Values
Visual Structures
Map data to a graphical form that encodes the
information
Three main properties: the spatial substrate,
markings, retinal properties
Main goal is effectiveness, which revolves
around perception
Spatial Substrate
Space is perceptually dominant
Spatial position is best method of visual
encoding
Four types of axes:
U = Unstructured Axis
N = Nominal Axis
O = Ordinal Axis
Q = Quantitative Axis
Spatial Substrate(2)
Several techniques to increase the amount of
information that can be encoded with spatial
position:
Composition
Alignment
Folding
Recursion
Overloading
Marks
Visible things that occur in space
Four elementary types of marks:
Connection and Enclosure
Points and lines can be used for topological
structure. I.E. Graphs and Trees
Trees and graphs can give different Gestalt
properties
Gestalt Properties
Gestalt’s Principles of organization:
Pragnanz
Proximity
Similarity
Closure
Good Continuation
Common fate
Familiarity
Retinal Properties
Bertin suggested six retinal properties:
View Transformations
Interactively modify and augment Visual
Structures
Three common view transformations:
Location Probes
Viewpoint Controls
Distortions
The Eyes Have It
A Task by Data Type Taxonomy for Information
Visualizations
Starting a Visualization
Most visualizations start off with the following
mantra: Overview first, zoom and filter, then
details on demand
Not the only way to start in information
visualization.
Many different data types and tasks for
visualizations
Data Types
In Information Visualization, there are seven
diferent data types:
1-,2-,3-D data
Temporal data
Multi-dimensional data
Tree and network data
1-D Data
Linear data types
Items organized in a sequential manner
Interface issues are usually in finding good
markings and how to view
Example: bifocal display
2-D Data
Planar or map data
Each item has task-domain attributes and
interface-domain features
User problems are usually to find adjacent
items, containment, paths, counting, filtering,
details-on-demand
Example: GIS
3-D Data
Items with volume and potentially complex
relationships
Users tasks deal with adjacency, comparative
relationships
Users must cope with orientation, positioning
problems when viewing data
Solutions include overviews, landmarks,
perspective, stereo display, transparency and
color coding
Temporal Data
Data that varies over time. Items have a start
and finish time.
Time lines widely used
Example: project management tools
Multi-Dimensional Data
Data in which n attributes become points in ndimensional space
Typical tasks include finding patterns, clusters,
correlations among pairs of variables, gaps,
and outliers
Parallel coordinates is a clever innovation
Example: HomeFinder
Tree Data
Collections of items with each item linking to a
parent item
Interface representations include node and
link diagram, treemap
Example: TreeBrowser
Network Data
Data items linked to an arbitrary number of
items
Users want to know best path, how to traverse
network
Interface representations include node and
link diagrams and square matrices.
Visualization Tasks
There are seven different types of visualization
tasks:
Overview
Zoom
Filter
Details-on-Demand
Relate
History
Extract
Overview
Gain overview of entire collection
Strategies include different zoomed views of
individual data items, with detail view
Popular approach is Fisheye
All Information Visualization interfaces support
some overview strategy
Zoom
Zoom in on items of interest
User typically wants to see one particular
region of data
Satisfying method is by clicking and zooming
Filter
Filter out unwanted items
With good content tools, users can quickly
focus on areas of interest.
Rapid update to sliders, buttons, etc is
important
Details on Demand
Select individual data element and gain more
information
Usually done after filtering or zooming to fewer
items
Typical approach is clicking on an item and a
pop up appears
Relate
View relationships among items
Example: FilmFinder comparing directors
History
Keep history of actions to enable undo,
refinement operations
Most designers ignore this
Rare that single user item produces a desired
outcome
Extract
Allow extraction of sub-collections and of the
query parameters
Save to database or file in a new format
Most prototypes ignore this
Conclusions
We have covered two important parts of
Information Visualization:
What Infovis is, and how to create a
representation
How to provide interactive tools for InfoVis and
the different types of data