Value-Suppressing Uncertainty Palettes Michael Correll Dominik - - PowerPoint PPT Presentation

▶
value suppressing uncertainty palettes
SMART_READER_LITE
LIVE PREVIEW

Value-Suppressing Uncertainty Palettes Michael Correll Dominik - - PowerPoint PPT Presentation

Nov 30, 2023 286 likes •1.32k views

Value-Suppressing Uncertainty Palettes Michael Correll Dominik Moritz Jeffrey Heer 1 Value-Suppressing Uncertainty Palettes Michael Correll Dominik Moritz Jeffrey Heer 2 Outline What We Did Why We Did It Why We Think It Works 3

slide-1
SLIDE 1

Value-Suppressing Uncertainty Palettes

Michael Correll Dominik Moritz Jeffrey Heer

1

slide-2
SLIDE 2

Value-Suppressing Uncertainty Palettes

Michael Correll Dominik Moritz Jeffrey Heer

2

slide-3
SLIDE 3

Outline

What We Did Why We Did It Why We Think It Works

3

slide-4
SLIDE 4

Outline

What We Did Why We Did It Why We Think It Works

4

slide-5
SLIDE 5

2016 Election

5

slide-6
SLIDE 6

2016 U.S. Election

6

slide-7
SLIDE 7

2016 U.S. Election

Clinton is only 1.6 margins

  • f error away! (p>0.05)

7

slide-8
SLIDE 8

2016 U.S. Election

Clinton is only 1.6 margins

  • f error away! (p>0.05)

How to convey uncertainty?

8

slide-9
SLIDE 9

2016 U.S. Election

Clinton is only 1.6 margins

  • f error away! (p>0.05)

How to convey uncertainty? How to encourage people from refraining from deciding?

9

slide-10
SLIDE 10

Value-Suppressing Uncertainty Palette

10

slide-11
SLIDE 11

Value-Suppressing Uncertainty Palette

11

slide-12
SLIDE 12

Value-Suppressing Uncertainty Palette

12

slide-13
SLIDE 13

Value-Suppressing Uncertainty Palettes

Traditional Bivariate Map

13

slide-14
SLIDE 14

Value-Suppressing Uncertainty Palettes

Traditional Bivariate Map VSUP

14

slide-15
SLIDE 15

Value-Suppressing Uncertainty Palettes

Traditional Bivariate Map VSUP

15

slide-16
SLIDE 16

Value-Suppressing Uncertainty Palettes

Traditional Bivariate Map VSUP

16

slide-17
SLIDE 17

Value-Suppressing Uncertainty Palettes

Traditional Bivariate Map VSUP

17

slide-18
SLIDE 18

VSUP

18

slide-19
SLIDE 19

Tree-Quantization

19

slide-20
SLIDE 20

Tree-Quantization

20

slide-21
SLIDE 21

Tree-Quantization

21

slide-22
SLIDE 22

22

slide-23
SLIDE 23

Outline

What We Did Why We Did It Why We Think It Works

23

slide-24
SLIDE 24

Uncertainty In Maps

24

slide-25
SLIDE 25

Uncertainty In Maps

Data Map

25

slide-26
SLIDE 26

Uncertainty In Maps

Data Map Uncertainty Map

26

slide-27
SLIDE 27

Uncertainty In Maps

Data Map Uncertainty Map

27

slide-28
SLIDE 28

Uncertainty In Maps

Data Map Uncertainty Map

28

slide-29
SLIDE 29

Uncertainty In Maps

Data Map Uncertainty Map

29

slide-30
SLIDE 30

Uncertainty In Maps

Data Map Uncertainty Map

30

slide-31
SLIDE 31

Juxtaposition

Data Map Uncertainty Map

31

slide-32
SLIDE 32

Data Map

32

slide-33
SLIDE 33

Juxtaposed Uncertainty

33

slide-34
SLIDE 34

Why This Is Bad

2x Space Requirements

34

slide-35
SLIDE 35

Why This Is Bad

2x Space Requirements Requires Searching

35

slide-36
SLIDE 36

Why This Is Bad

2x Space Requirements Requires Searching Ignorable

36

slide-37
SLIDE 37

Superposition

37

slide-38
SLIDE 38

Bivariate Map

1928

38

slide-39
SLIDE 39

Bivariate Map

1928 1874

39

slide-40
SLIDE 40

40

slide-41
SLIDE 41

VSUP

41

slide-42
SLIDE 42

VSUP

VSUPs are a bivariate mapping of data and uncertainty that allow fine-grain comparisons when data are more certain, and coarser comparisons when data are less certain.

42

slide-43
SLIDE 43

Outline

What We Did Why We Did It Why We Think It Works

43

slide-44
SLIDE 44

Will This Work?

44

slide-45
SLIDE 45

Will This Work?

45

slide-46
SLIDE 46

Bivariate Maps Are Hard!

“[R]eading Two-Variable Color Maps at the elementary, intermediate, or superior level is at the very least difficult, and may be impossible.”

  • Wainer & Francolini, 1980

46

slide-47
SLIDE 47

Discrete Color Maps Are Inaccurate!

“[W]ith possible rare exceptions, continuous color scales represent the data more effectively than binned color scales, so we should stick with them.”

  • Few, 2017

47

slide-48
SLIDE 48

Evaluating VSUPs

48

slide-49
SLIDE 49

Evaluating VSUPs

49

slide-50
SLIDE 50

Evaluating VSUPs

50

slide-51
SLIDE 51

Alternative Designs

51

slide-52
SLIDE 52

Alternative Designs

52

slide-53
SLIDE 53

Alternative Designs

53

slide-54
SLIDE 54

Alternative Designs

54

slide-55
SLIDE 55

Alternative Designs

55

slide-56
SLIDE 56

Alternative Designs

56

slide-57
SLIDE 57

Methods

2 MTurk Experiments: Identification Task Prediction Task 48 Participants (w/o CVD) 2112 Trials

57

slide-58
SLIDE 58

Identification Task

58

slide-59
SLIDE 59

Value of 0.45, Uncertainty of 0.1

59

slide-60
SLIDE 60

Value of 0.45

60

slide-61
SLIDE 61

Value of 0.45

61

slide-62
SLIDE 62

Uncertainty of 0.1

62

slide-63
SLIDE 63

Uncertainty of 0.1

63

slide-64
SLIDE 64

Uncertainty of 0.1

64

slide-65
SLIDE 65

Identification Task

65

slide-66
SLIDE 66

Value of 0.7, Uncertainty of 0.4

66

slide-67
SLIDE 67

Uncertainty of 0.4

67

slide-68
SLIDE 68

Uncertainty of 0.4

68

slide-69
SLIDE 69

Value of 0.7

69

slide-70
SLIDE 70

Value of 0.7

70

slide-71
SLIDE 71

Value of 0.7

71

slide-72
SLIDE 72

Results

72

slide-73
SLIDE 73

Results

73

slide-74
SLIDE 74

Value of 0.7, Uncertainty of 0.4

74

slide-75
SLIDE 75

Value of 0.7, Uncertainty of 0.4

75

slide-76
SLIDE 76

Results

76

slide-77
SLIDE 77

Results

77

slide-78
SLIDE 78

Results

78

slide-79
SLIDE 79

Results

Interference disrupts Continuous Maps Search disrupts Juxtaposed Maps

79

slide-80
SLIDE 80

Prediction Task

How can we induce and compare risk-averse behavior in heatmaps?

80

slide-81
SLIDE 81

Prediction Task

81

slide-82
SLIDE 82

Prediction Task

82

slide-83
SLIDE 83

Prediction Task

83

slide-84
SLIDE 84

Prediction Task

84

slide-85
SLIDE 85

Prediction Task

Could be Dangerous! Definitely Safe.

85

slide-86
SLIDE 86

Placing Ships

86

slide-87
SLIDE 87

Where To Put The Last Ships?

87

slide-88
SLIDE 88

Where To Put The Last Ships?

88

slide-89
SLIDE 89

Where To Put The Last Ship?

89

slide-90
SLIDE 90

Where To Put The Last Ships?

90

slide-91
SLIDE 91

Where To Put The Last Ships?

91

slide-92
SLIDE 92

Results

Bivariate Maps encourage risk seeking VSUP Maps encourage risk aversion

92

slide-93
SLIDE 93

VSUPs are an unignorable way of integrating data and uncertainty. VSUPs make people more cautious in their decision-making.

93

slide-94
SLIDE 94

Thanks!

This work was supported by a Moore Foundation Data-Driven Discovery Investigator award. Study materials available at: https://github.com/uwdata/papers-vsup Make your own VSUPs at: https://github.com/uwdata/vsup

94

slide-95
SLIDE 95

Extra Slides

95

slide-96
SLIDE 96

96

slide-97
SLIDE 97

Flight Delay

97

slide-98
SLIDE 98

Flight Delay

98

slide-99
SLIDE 99

Flight Delay

99

slide-100
SLIDE 100

100

slide-101
SLIDE 101

Can you do this in Tableau?

101

“How to make effective bivariate choropleth maps with Tableau” Sarah Battersby https://www.tableau.com/about/blog/ 2018/2/how-make-effective-bivariate- choropleth-maps-tableau-83121

slide-102
SLIDE 102

Have you Used this for Anything?

102