The Process of Visual Thinking Colin Ware University of New Hampshire Ware:Vislab:CCOM Visual Thinking Virtual Machine     Capture common interactive processes Analytic tools for designers Can be.

Download Report

Transcript The Process of Visual Thinking Colin Ware University of New Hampshire Ware:Vislab:CCOM Visual Thinking Virtual Machine     Capture common interactive processes Analytic tools for designers Can be.

The Process of Visual Thinking
Colin Ware
University of New Hampshire
Ware:Vislab:CCOM
Visual Thinking Virtual Machine




Capture common interactive processes
Analytic tools for designers
Can be partially automated (recommenders)
Based on a virtual machine
Three channels
elements
of f orm
form
mov ement
color
1
3
2& A
4 B
4
C
s 4
r
5
ye
a
6
l
l
a
c
rti
o
c
color
color
dif f erences
elements
of mov ement
Adapted from Livingston and Hubel
Rapid propagation along contours
Theory: Field, Hayes and Hess, 1998
Theory applied to flow visualization: Visualizations producing the
most activation of Cortical neurons with receptive fields tangential to
the flow direction will perform the best.
2D Flow visualization
• A landmark study Laidlaw et al (2001)
Task:
Advection
perception
Modeling V1 and above
Pineo and Ware, 2011
How to add direction along
contour?
Halle’s “little stroaks” 1868
Asymmetry along path
End-Stopped neurons
Fowler and Ware (1988)
Most flow current and wind
displays are poor.
Ware:Vislab:CCOM
Ware:Vislab:CCOM
Visual working memory
Ware:Vislab:CCOM
Ware:Vislab:CCOM
Ware:Vislab:CCOM
Ware:Vislab:CCOM
Ware:Vislab:CCOM
Ware:Vislab:CCOM
The visual query
 Transforming a problem into a pattern
search
 E.g. path in a network diagram
More visual queries
Vowel formants
Ware:Vislab:CCOM
Epistemic actions
 Actions executed to seek
knowledge
 E.g. an eye movement. 70
msec
 Or – brushing, dynamic queries,
zooming
 The essence of interactive
visualization
Epistemic actions
Example 1: VT Process of
Discovering novel patterns in time
varying data
 Tools for finding new underwater
behaviors from humpback whale tag data
(Why turning time into space is a good idea)
The gear
Antenna
DTAG Mark Johnson
Big
Eyes
Dave Wiley
Task: find new behaviors
 = stereotyped patterns
Visual thinking Process
 Repeat
• Review behavior sequence looking for
patterns. Remember patterns.
• Look for more instances.
 until no new patterns
Solution 1. GeoZui 4D
Cognitive process for finding new
behaviors
 = stereotyped patterns
Visual Thinking Process
 repeat
• Review behavior sequence looking for patterns by
playback. Remember patterns using space-time
notes.
• Look for more instances. May involve reviewing all
other whale tracks.
 Until no new patterns
 Cost k*playback time.
Solution 2: trackplot
Foraging patterns
Traversing
2006
Mostly
04
06
07
06
Process for finding new
behaviors
Cognitive algorithm
 Get to a good viewpoint
 repeat
• Review behavior sequence looking for patterns eye
movements. Remember patterns using visual working
memory.
• Look for more instances. May involve reviewing all
other whale tracks. Can be posted on the wall
 until no new patterns
 Cost Nav + Eye Movement time *pattern
matching.
Gain in efficiency – from
playback tool to pattern finding
tool
 Many hours (with playback)
 A few minutes or days (with patterns)
 Approximately a factor of 100
Example 2. VT process with degree of
relevance highlighting
ME Graph (Ware and Bobrow)
Degree of relevance higlighting
Process:Degree of relevance
highlighting
1. Construct visual query to find a symbol that may lead to
useful information (information scent).
2. Execute an epistemic action by selecting a symbol.
3. Computer highlights all symbols with a high degree of
relevance to selected symbol.
4. Execute a visual pattern query among highlighted
symbols for additional information scent.
5. If a very high relevance symbol is found, execute an
epistemic action to drill down for additional information.
Usually this will be presented in a different display
window.
6. Repeat from 1 as needed, cognitively marking visited
Ware:Vislab:CCOM
The Design Process
Cognitive
Task analysis
+ Data
affordances
Design
With Aperture
VTDP display budgets
Small network (<30 nodes) – static
representation
Medium graph (>30,<600) - use Degree of
Relevance VTDP
Large Graph (>600 < 5m) use
Dynamic queries VTDP on
donors and charity attributes
Recommenders:
Interaction
(VTDPs)
+ Data mappings
Prototypes
For evaluation
Cognitive
walkthoughs
Analyst
feedback
Product
Agile Visualization
Design for Cognitive Efficiency
Cookbook:
Details on Demand
ACT-R
VTDPs
Shneiderman
Anderson
Funding:
DARPA
NSF
NOAA
Ware:Vislab:CCOM
Channels have sub-channels
Visual separation between channels
Visual Interference within channels
Example 3: Process of searching for small
patterns in a large information space
Zooming vs
Windows + eye movements
-A formal study
Prediction
> VWM capacity
- Add extra windows
Results
Tight coupling
 Supports display of < 500 items
 Order of magnitude improvement
 Epistemic action rate can be one per
second
Ware:Vislab:CCOM
Cognitive Model (grossly
simplified)
 Time = setup cost +
number of “visits” x time per visit
 Number of “visits” is a function of number of
objects to be compared and visual working
memory capacity.
visits = n/M
 Results:extra window much more efficient if
working memory is exceeded
 Errors drop enormously