RubyConf_2017_-_Kelsey_Pedersen.pdf

Transcript

1. HUMAN DECISION MAKING Kelsey Pedersen Software Engineer, Stitch Fix with DATA SCIENCE AUGMENTING @kelsey_pedersen
2. @kelsey_pedersen KELSEY PEDERSEN
3. I’m a software engineer at Stitch Fix - a personalized styling service
10. stylists data scientists
13. How do humans make decisions? What are the limitations of human decision making? How can we help users make decisions with data science?
14. How do humans make decisions? Human Decisions Limits of Decisions Stylists Decisions Humans & Data Science The Future
15. System 1 System 2
16. System 1 impressions & feelings
21. 95% of human decisions are made instantaneously
22. System 2 analytical & effortful
23. 1737 x 1990 Answer: 3,456,630
26. System 1 gut feelings System 2 computations
27. LEFT BRAIN RIGHT BRAIN
28. What are the limits in human decision making? Human Decisions Limits of Decisions Stylists Decisions Humans & Data Science The Future
30. Human decisions are unpredictable
31. Inconsistent Judgments
32. Inconsistent Judgments
33. Inconsistent Judgments
34. Human decisions are driven by your own past experiences
35. Limited Information
36. Limited Information
37. Limited Information
38. Human decisions are driven by your own views and preferences
39. Biased Decisions
40. Biased Decisions
41. Biased Decisions
42. RISKY & UNPREDICTABLE
43. What are examples of Stitch Fix stylists making decisions? Human Decisions Limits of Decisions Stylists Decisions Humans & Data Science The Future
44. Inconsistent Judgments by Stylists
45. Inconsistent Judgments by Stylists
46. Limited Information by Stylists
47. Limited Information by Stylists
48. Biased Decisions by Stylists
50. In what ways can data science augment human decisions? Human Decisions Limits of Decisions Stylists Decisions Humans & Data Science The Future
52. Data Science Noun: Use of mathematics or statistics to answer a business question
53. GUIDE DECISIONS WITH COMPUTATIONS TRAIN DECISIONS WITH FEEDBACK
54. DATA STYLING STYLING STYLING STYLING STYLING STYLING STYLING STYLING DATA DATA DATA DATA DATA
56. Deciding on one item
57. GUIDE Predictive algorithms
58. 72% 60% 48% 44% GUIDE Predictive algorithms
59. GUIDE Predictive algorithms
60. GUIDE Predictive algorithms
62. Deciding on all 5 items
63. 72% 60% 48% 44% GUIDE Predictive algorithms 76%
64. 72% 60% 48% 44% GUIDE Predictive algorithms 24%
66. Client Feedback on all 5 items
70. Client Feedback overtime for one Stylist
71. TRAIN Performance Metrics
72. TRAIN Performance Metrics
74. All Feedback for all Stylists
77. Can humans augment algorithms? Human Decisions Limits of Decisions Stylists Decisions Humans & Data Science The Future
83. 72% 60% 48% 10%
84. 72% 60% 48% 44% 24%
87. stylists data scientists
88. What is the future of data science? Human Decisions Limits of Decisions Stylists Decisions Humans & Data Science The Future
90. LEFT BRAIN RIGHT BRAIN VS.
91. System 1 gut feelings System 2 computations &
94. Data Science Humans System 1 System 2 Data Science Humans System 3
95. System 3 predictive & intuitive
96. stylists data scientists
97. TRAIN Performance Metrics
99. System 1 gut feelings System 2 computations System 3 feedback
100. What is the business impact of system 3?
101. lower labor cost fewer mistakes from humans greater client satisfaction increased keep rate
102. In conclusion…
103. Conclusion
106. Conclusion
107. humans data science
108. @kelsey_pedersen KELSEY PEDERSEN