Visual (Mis)perceptions
stat/engl 332
‘Chart is interesting, if we
see something interesting’
• pre-attentive tasks are rapid and ‘pop’ from
our visual field
• if charts use pre-attentive features,
audience does not need to ‘think first’
before looking at a chart, attention drawing
• pre-attentive tasks: target detection,
boundary detection, ‘counting’, tracking
Target Detection
•Absence/Presence of Target
•Preattentive features: color, shape
Absent
Present
http://www.csc.ncsu.edu/faculty/healey/PP/
Change Blindness
Interruption (Blink, Eye Saccade, Blank Screen) makes
significant changes “invisible” - until attention is focused
directly on the object
http://cognitrn.psych.indiana.edu/CogsciSoftware/ChangeBlindness/
Change Blindness
Interruption (Blink, Eye Saccade, Blank Screen) makes
significant changes “invisible” - until attention is focused
directly on the object
http://cognitrn.psych.indiana.edu/CogsciSoftware/ChangeBlindness/
Conclusions so far
visual memory is very limited, tiny disruptions (blink of
the eye) cause “loss of data”
!
“spot the difference” type charts are hard to digest (e.g.
star plots, Chernoff faces)
resolution: render information itself rather than
expecting audience to derive it
Visual System
very well adapted to “standard” situations - “hard-wired”
artificial manipulations cause inappropriate
interpretations of the visual scene
idea: use visual illusions to find out about perception
Face Detection
Mooney Faces
Face in the Beans
Orientation of Faces
Adaption
Nerve cells fatigue after responding for
several seconds to the same stimulus:
e.g. stare at bright red cross for several
seconds....
Nerve cells fatigue after responding for
several seconds to the same stimulus:
e.g. stare at white cross for several seconds....
!
.. cross appears briefly in complementary
color on turquoise background.
Motion Aftereffect
•Example: Buddha
Motion
•Perceived Movement depends on contrast
•e.g. Rotating Snake w/o light, stepping feet
Power of Context
•Which line is longer?
Context
“angles in” (near the ticket
counter) is front side of
object,
“angles out” occurs at the far
end of a room
given identical retinal size brain assumes the “angles in” configuration to be closer, and
concludes that the “angle in”-line is shorter.
Power of Context
Are the grey values of A and B different?
Power of Context
Describe the line lengths
Biological Motion
Seemingly unconnected dots are
placed into context by motion
Motion Induced Blindness
Focus on green spot in the
middle - the yellow spots will
disappear ... singly, in pairs
or all three simultaneously
Sustained inattentional
Blindness
•Selective Looking (Simons, Chabris 99)
•http://www.simonslab.com/videos.html
Application to
Graphical Elements
Stat/Engl 332
Goal
Evaluate perceptual strengths and weaknesses
Example: Bar vs Pie
•What tasks are involved?
Pie Chart of some Fake Data
Barchart of same Fake Data
6
5
45
1
4
2
1
2
3
4
5
6
40
35
30
25
20
15
10
5
0
1
3
2
3
4
5
6
Area is proportional to value
!
comparison of angles, curve length
comparison of heights, positions along a common scale
Ranking of perceptual tasks
• usually we are not interested in exact quantities
• ... But ... use accuracy as measure
Cleveland & McGill premise:
!
A graphical form that involves elementary perceptual
tasks that lead to more accurate judgments than
another graphical form (with the same quantitative
information) will result in a better organization and
increase the chances of correct perception of patterns
and behavior.
Evaluation of different designs
User studies!
Positions along a common scale
Determine the values for bins 1 to 6 as
accurately as possible
Bin
1
2
3
4
5
6
Barchart
30
25
20
15
10
5
0
1
2
3
4
5
Value
5
27
9
25
20
14
6
write down differences between
true values and your estimates
Angle comparisons
Determine
thethe
angles
for slices
to 6 as
Determine
percentage
for1slices
1
accurately
as possibleas possible
to 6 as accurately
Pie Chart
6
5
1
2
3
1
4
5
6
4
3
Slice Value
29
1
9
2
10
3
27
4
11
5
13
6
2
write down differences between
true values and your estimates
Evaluating User Perception
Get differences between user estimates
and true values
Square differences
Sum all squared differences
Divide sum by #differences-1
(= average error)
Results & Evaluation
• Graphically?
• “Statistically”?
!
• What would also be
interesting?
Error
0
1
2
3
4
5 or more
Bar
Pie
Positions along non-aligned scales
Determine the values of B and C in comparison to A as accurately as possible
B = 0.7A
C = 1.1A
A
B
C
Comparison of Areas
Determine the values of B and C in comparison to A as accurately as possible
B = 0.5A
C = 2A
A
B
C
Volume Comparisons
3d Exploded Pie Chart
6
1
5
2
3
4
1
2
3
4
5
6
Ranking of Difficulty of Graphical Tasks
(Cleveland & McGill)
•
•
•
•
•
•
Position along a common scale
Position along non-aligned scales
Length, Direction, Angle
Area
Volume, Curvature
Shading, Color Saturation