Machine Learning for Designers

Scott Sullivan, LPK
ProfsoUX 2019

The presentation is targeted for mid-level to advanced human-centered designers, who want to meaningfully contribute to projects that involve Machine Learning.

AI is arguably the most significant advancement in technology since the internet, it’s currently transforming everything we touch from text translation, to the core architecture of Netflix’s entire product offering. In the rapidly changing competitive business landscape, there are massive opportunities for the automation of services that used to be labor intensive, but beyond automation, there is a new frontier of innovative services, that have no resemblance to current digital offerings.

In his new talk, Scott Sullivan, Director of User Experience at LPK, will give an overview of what’s currently possible with Machine Learning and Artificial Intelligence, and outline the outcome-based opportunities of the emerging technology. Beyond the basics, Scott will outline UX’s role and relationship with Machine Learning/AI, and delve in to the ethics and pitfalls of AI and behavior change.

Transcript

1. Scott Sullivan @scotsullivan lpk.com Machine Learning for Designers
5. WHAT IS MACHINE LEARNING?
6. WHAT IS MACHINE LEARNING? “improving performance in some task with experience” -Tom Mitchell, author of Machine Learning
7. WHAT IS MACHINE LEARNING? “it is a method of teaching computers to make and improve predictions or behaviors based on some data.” -Daniel G, guy on Stack Overflow
8. WHAT IS MACHINE LEARNING? “it is a method of teaching computers to make and improve predictions or behaviors based on some data.” -Daniel G, guy on Stack Overflow
9. WHAT IS MACHINE LEARNING? What’s different?
10. WHAT IS MACHINE LEARNING? What’s different? traditional code: Explicit (instructions)
11. WHAT IS MACHINE LEARNING? What’s different? traditional code: Explicit machine learning: Implicit (instructions) (train with data)
12. WHAT IS MACHINE LEARNING? Software that is written by showing it data. Then it makes predictions. ! Our Data Model Small part of software
13. WHAT IS MACHINE LEARNING? Classification
14. WHAT IS MACHINE LEARNING? Regression
15. WHAT IS MACHINE LEARNING? Clustering
16. WHAT IS MACHINE LEARNING?
17. PROCESS / SKETCHING
18. PROCESS / SKETCHING
19. PROCESS / SKETCHING
20. MICRO EXERCISE 1: PSEUDOCODING
21. MICRO EXERCISE 1: PSEUDOCODING PSEUDOCODE WHAT 1 2 3 4 5 6 7 8 9 10 WHAT 1 2 3 4 5 6 7 8 9 10 • Describe step-by-step how to brush your teeth on the left side of the page. 5 minutes Brush your teeth Pick up your tooth brush
22. MICRO EXERCISE 1: PSEUDOCODING PSEUDOCODE WHAT 1 2 3 4 5 6 7 8 9 10 WHAT 1 2 3 4 5 6 7 8 9 10 • Describe step-by-step how to recognize a human face. 5 more minutes Brush your teeth Pick up your tooth brush Recognize a face
23. MACHINE LEARNING AS DESIGN MATERIAL
24. MACHINE LEARNING AS DESIGN MATERIAL
25. MACHINE LEARNING AS DESIGN MATERIAL
26. MACHINE LEARNING AS DESIGN MATERIAL
27. MACHINE LEARNING AS DESIGN MATERIAL
28. MACHINE LEARNING AS DESIGN ARCHITECTURE
29. MACHINE LEARNING AS DESIGN ARCHITECTURE 8 hours of sleep 7.5 hours of sleep 7 hours of sleep 6.5 hours of sleep 6 hours of sleep 5.5 hours of sleep TARGET VARIABLE INPUT VARIABLES
30. MACHINE LEARNING AS DESIGN ARCHITECTURE 8 hours of sleep 7.5 hours of sleep 7 hours of sleep 6.5 hours of sleep 6 hours of sleep 5.5 hours of sleep goal current state
31. MACHINE LEARNING AS DESIGN ARCHITECTURE 6 5.5
32. MACHINE LEARNING AS DESIGN ARCHITECTURE 6 5.5
33. MACHINE LEARNING AS DESIGN ARCHITECTURE 6 5.5
34. MACHINE LEARNING AS DESIGN ARCHITECTURE 6 5.5
35. MACHINE LEARNING AS DESIGN ARCHITECTURE 6 5.5 Activity Time went to bed Alcohol consumption Time spent at bar
36. MACHINE LEARNING AS DESIGN ARCHITECTURE 6 5.5 Activity Time went to bed Alcohol consumption Time spent at bar
37. MACHINE LEARNING AS DESIGN ARCHITECTURE 6 5.5 Activity Time went to bed Alcohol consumption Time spent at bar
38. MACHINE LEARNING AS DESIGN ARCHITECTURE 6 5.5 Activity Time went to bed Alcohol consumption Time spent at bar
39. PROCESS
40. Designers need to do this. ! Our Data Model Small part of software
41. PROCESS / HANDS ON
42. PROCESS / HANDS ON WITH DATA Tableau
43. PROCESS / SKETCHING
44. PROCESS / SKETCHING
45. PROCESS / SKETCHING
46. PROCESS / IDEATION
47. PROCESS / IDEATION
48. PROCESS / IDEATION • What are the goals? • What are the input variables? • What is the target variable? If it was a human, how would you tell them how to do it.
49. PROCESS / IDEATION • What are the goals? • What are the input variables? • What is the target variable? • What output are you expecting? If it was a human, how would you tell them how to do it.
50. If it was a human, how would you tell them how to do it. PROCESS / IDEATION • What are the goals? • What are the input variables? • What is the target variable? • What output are you expecting?
51. PROCESS / IDEATION • What are the goals? • What are the input variables? • What is the target variable? • What output are you expecting? If it was a human, how would you tell them how to do it.
52. ME3: IDEATION TARGET VARIABLE GOAL INPUT VARIABLES EXPECTED OUTPUTS • What’s your goal? • What’s your target variable? • What are your input variables? • what are your expected outputs? If it was a human, how would you tell them how to do it.
53. PROCESS / DIALOG
54. PROCESS / DIALOG “… the center of design becomes the system and it’s outcomes. Design moves towards building emergent ecologies” - Philip Van Allen, ArtCenter College of Design
55. PROCESS / DIALOG EXPLICIT FEEDBACK • Goal setting • Preferences • Answering questions • Manual Adjustments to their model USER INPUT
56. PROCESS / DIALOG
57. PROCESS / DIALOG
58. PROCESS / DIALOG EXPLICIT FEEDBACK • Goal setting • Preferences • Answering questions • Manual Adjustments to their model USER INPUT IMPLICIT FEEDBACK • Adjustments to behavior • Changes in engagement
59. PROCESS / DIALOG TRANSPARENCY • Demonstrate the decision provenance in the structure of the dialog • Say where you got the data • Show a confidence score (classification) • Show the next couple guesses (classification) • Don’t speak in absolutes
60. PROCESS / DIALOG
62. DIALOG / PROTOTYPING
63. DIALOG / PROTOTYPING
64. DIALOG / PROTOTYPING
65. DIALOG / ∞
66. DIALOG / ∞ IMPROVEMENT How would you tell a human how to improve the predictions? FEEDBACK AWARENESS What assumptions should the model not make?
67. DIALOG / ∞ BIAS TYPES • Input bias, in your source data • Introduced bias, in your algorithm • Learned bias, learned by your model BIAS AWARENESS • What is the system learning? • What is it learning from? • How is it using this knowledge in it’s decisions?
68. THANK YOU! Scott Sullivan LPK @scotsullivan LPK.com
69. MIT LICENSE Copyright 2019 Scott Sullivan Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.