Spatial Displays - PowerPoint PPT Presentation

1 / 77
About This Presentation
Title:

Spatial Displays

Description:

So, if circles are used to graph values, therefore, observers will underestimate ... or distant car? Object Centered cues. Texture gradients ... – PowerPoint PPT presentation

Number of Views:54
Avg rating:3.0/5.0
Slides: 78
Provided by: hfac
Category:
Tags: displays | spatial

less

Transcript and Presenter's Notes

Title: Spatial Displays


1
Spatial Displays
  • Lecture 4

2
Optimal .493
Observed .24
Signal
Noise
P (X N or SN)
X (Internal Signal)
3
Spatial/Analog Perception perception of
continuous characteristics relationships
distinguished from perception of digital stimulus
characteristics For example, perception of
length perception of angle vs. perception
of words perception of digits
Attitude 49 mph Altitude 99
mph Height 101 mph
4
With analog representation, magnitude of physical
difference indicates magnitude of difference in
meaning. With digital representation, magnitude
of physical difference does not correspond to
magnitude of difference in meaning. Creating
Graphs Graph static analog representation of
multiple numeric data points
5
Graphing Data 1. Avoid perceptual biases 2.
Consider the task/Minimize mental operations 3.
Code multiple graphs consistently 4. Keep
data-to-ink ratio high 5. Code multiple graphs
consistently
6
Poggendorf Illusion
7
Tends to make lines look flatter than they are
That is, without context, people tend to
underestimate the slope.
8
(No Transcript)
9
(No Transcript)
10
A crime against aesthetics
11
Estimating differences with changing slopes
12
Estimating differences with changing slopes
13
Estimating differences with changing slopes
Difference score might be more informative
14
Creating Graphs Avoid perceptual biases Be
aware of tendency to overestimate steepness of
regression lines
15
Creating Graphs Avoid perceptual biases Be
aware of tendency to overestimate steepness of
regression lines
16
  • Creating Graphs
  • Avoid perceptual biases
  • Code numerical values using an ordered continuum
  • brightness
  • length
  • not color (it is not an ordered continuum)
  • be aware of response compression expansion

17
(No Transcript)
18
Area a
Area 4a
Area 9a
Area .25a
19
Avoid perceptual biases Perception of magnitude
often systematically biased Stevens power law
P k Ia P perceived magnitude I
Physical magnitude a Stevens exponent
K constant Response compression occurs when
exponent is less than 1 perceived magnitude
increases more slowly than physical
magnitude Response expansion occurs when
exponent is greater than 1 perceived magnitude
increases faster than physical magnitude
20
Creating Graphs Avoid perceptual biases Be
aware of response compression expansion if
exponent for a stimulus dimension does not equal
1, people may misestimate relations between
values graphed with that dimension For example,
Because the exponent for area is less than 1,
observers will underestimate the difference in
area between two circles. So, if circles are
used to graph values, therefore, observers will
underestimate difference in values being
portrayed.
21
Expansion
Unbiasedness
Compression
22
Dimension Exponent Visual Area 0.7 (compre
ssion) Visual Length 1.0 Brightness 1.0 Dura
tion 1.1 (expansion) Reflectance of gray
paper 1.2 (expansion) Redness (color
saturation) 1.7 (expansion) Electric
shock 3.5 (expansion) Electric shock to
teeth 7.0 !!!!!
23
Creating Graphs Avoid perceptual biases To
mitigate effects of response compression
expansion graph values using stimulus
dimension with Stevens exponent 1 (e.g.,
length) increase the frequency of tickmarks on
the graph - tickmarks should be on the inside,
if possible
24
  • Creating Graphs
  • If the goal is precise graph reading, stay away
    from fine graduations of color.
  • Do not use more than 5 or so colors.

25
(No Transcript)
26
Graphing Data 1. Avoid perceptual biases 2.
Consider the task/Minimize mental operations 3.
Code multiple graphs consistently 4. Keep
data-to-ink ratio high 5. Code multiple graphs
consistently
v
27
Consider the Task/Minimize Mental
Operations Taxonomy of Graph Reading
Tasks Point readingdetermine the value of a
single data point Local comparisoncompare
values directly shown on the graph Global
comparisoncompare values derived from values
shown Synthesis judgmentmake a general judgment
about pattern in full set of data Note that
tasks are ordered by increasing demand for data
integration. Proximity Compatibility Principle
tells us that display proximity should increase
as demand for data integration increases.
28
Consider Task/Minimize Mental Operations
Reaction Time
Display Size
Point ReadingHow fast were subjects in the
display size 5 control condition? Local
comparisonWere subjects faster in the
experimental condition or the control condition
for display size 2?
29
Consider Task/Minimize Mental Operations
Reaction Time
Display Size
Global comparisonWas the difference between
display sizes 1 and 3 bigger than that between
display sizes 3 and 5 for the Experimental
condition? Synthesis judgmentDo RTs for either
of the two conditions increase as a linear
function of display size?
30
Consider Task/Minimize Mental Operations
Control
Experimental
Reaction Time
Display Set Size
point readingHow fast were subjects in the
display size 4 control condition? Local
comparisonWere subjects faster in the
experimental condition or the control condition
for display size 2?
31
Consider Task/Minimize Mental Operations
Control
Experimental
Reaction Time
Display Size
Global comparisonWas the difference between
experimental and control conditions larger at
display size 2 or display size 4? Synthesis
judgmentDo RTs for either of the two conditions
increase as a linear function of display size?
32
Consider Task/Minimize Mental Operations
Reaction Time
Display Size
Global comparisonWas the difference between
experimental and control conditions larger at
display size 1 or display size 2? Synthesis
judgmentDo RTs for either of the two conditions
increase as a linear function of display size?
33
Consider Task/Minimize Mental Operations
Why does the Proximity Compatibility Principle
work? Note that various mental operations may be
involved in graph reading Mental scanning
mentally trace a path within the graph For
example, mentally trace an imaginary line from a
data point to the Y-axis, Or trace a line between
two bars in order to compare their heights
Mental imagery form a mental picture For
example, imagine a line connecting several bars
in order to determine if data are linear, Or
imagine one set of bars alongside another in
order to perform a global comparison
34
Consider Task/Minimize Mental Operations
What kinds of mental operations might be involved
in graph reading? Mental translation rotation
move or rotate a mental image For example,
rotate a pie graph to compare the sizes of two
wedges, Or mentally move a pair of bars next to
another pair in order to perform a global
judgment Attentional selection disregard
irrelevant info extract desired info For
example, estimate height of one data point
ignore others
35
Consider Task/Minimize Mental Operations
Why does the PCP work? Choice of representation
determines what mental operations are necessary
to extract a given type of info from a graph. To
facilitate ease of reading graphs, minimize the
number of mental operations necessary to derive
desired info. The PCP tends to minimize the
number of mental operations necessary to extract
desired info from a graph.
36
Consider Task/Minimize Mental Operations
Why does the PCP work? If info is high in
display proximity, mental operations are
necessary to isolate data points for local
judgments or local comparisons. If info is low
in display proximity, mental operations are
necessary to integrate data points for global
judgments or syntheses.
37
High proximity makes comparing lines
easy. However, high proximity makes extracting
individual points difficult.
38
Graphing Data 1. Avoid perceptual biases 2.
Consider the task/Minimize mental operations 3.
Code multiple graphs consistently 4. Keep
data-to-ink ratio high 5. Code multiple graphs
consistently
v
v
39
Code Multiple Graphs Consistently
When multiple graphs are necessary Use the
same stimulus feature to represent the same
variables Use the same feature values to
represent the same levels of a variable Note
that need for multiple graphs introduces Tradeoff
between local and global optimality Local
optimalityoptimality within a single
graph Global optimalityoptimality across a set
of graphs
40
(No Transcript)
41
(No Transcript)
42
Graphing Data 1. Avoid perceptual biases 2.
Consider the task/Minimize mental operations 3.
Code multiple graphs consistently 4. Keep
data-to-ink ratio high 5. Code multiple graphs
consistently
v
v
v
43
Maximize Data-to-Ink Ratio
Data-ink ratio ratio of amount of info conveyed
by graph to the amount of ink (lines, points,
text, other features) used to convey the info
eliminating unnecessary features generally
facilitates graph reading. Exceptions may arise
when features can help minimize mental
operations ink produces emergent features to
help minimize mental operations
44
(No Transcript)
45
Bad
Better
46
Better Lack of horizontal rules, but rules can
facilitate graph reading
Better
47
On the graph on the right, the addition of the
lines facilitates interpreting the interaction.
48
Pictorial Realism
  • Ideally, the display should mimic the physical
    layout and attributes of the data it is
    displaying.
  • Example altitude could be represented by an
    indicator that moves up and down.

49
6000
5000
4000
3000
2000
1000
0
50
  • Decreased range
  • must be large to have good resolution

6000
5000
4000
3000
2000
1000
0
51
  • circular motion does not mimic vertical motion
  • however, allows increased resolution in a small
    space.

52
800
  • unrealistic motion (numbers moving down climb)
  • Pictorial realism (high altitude at top)
  • realistic motion (numbers moving up climb)
  • lack of pictorial realism (high altitude at
    bottom)

700
600
500
400
300
200
100
53
  • motion not necessarily unrealistic.
  • altitude falls past plane as plane climbs.
  • unrealistic motion (numbers moving down climb)
  • Pictorial realism (high altitude at top)

54
Depth Perception
  • Object Centered cues characteristics of the
    outside world.
  • Observer-Centered cues characteristics of our
    visual system.

55
Object Centered cues
  • Linear Perspective

56
Object Centered cues
  • Interposition

57
Object Centered cues
  • Height in Plane

58
Object Centered cues
  • Light and Shadow

59
Object Centered cues
  • Familiar Size

or distant car?
Giant bike?
60
Object Centered cues
  • Texture gradients

texture becomes finer grained the closer it
appears
61
Object Centered cues
  • Aerial Perspective (distant objects are hazier)

62
Object Centered cues
  • Motion parallax (near items move more quickly)

y
c x
x o z
63
Object Centered cues
  • Motion parallax

y
c x
x o z
64
Object Centered cues
  • Motion parallax

y
c x
x o z
65
Object Centered cues
  • Motion parallax

y
c x
x o z
66
Object Centered cues
  • Motion parallax

y
c x
x o z
67
Observer-centered cues
  • Binocular disparity
  • the two eye view the world from slightly
    different angles

68
Observer-centered cues
  • Convergence
  • eyes turn inward to view near objects

69
Observer-centered cues
  • Accommodation
  • Lens changes shape to focus at different
    distances

70
Perceptual Hypotheses
  • we make an assumption about the state of the
    world based upon on the object-centered and
    observer-centered cues.
  • our hypothesis can be wrong when
  • too few clues
  • clues are ambiguous
  • we make the wrong assumptions

71
Perceptual Hypotheses
  • perceptual ambiguities are particularly likely to
    occur when representing the 3d world on a
    two-dimensional display.

72
3-D Graphs accurate reading requires accurate
depth size perception poor depth cues can
make depth size info ambiguous
73
(No Transcript)
74
3d representation of data
  • Because of the difficulty in making accurate
    depth judgments, 3d displays should be reserved
    for representing 3d worlds or objects.
  • However, the 3d visualization of data sets can
    help to reveal patterns that otherwise might be
    missed

75
(No Transcript)
76
(No Transcript)
77
(No Transcript)
Write a Comment
User Comments (0)
About PowerShow.com