Slide 1
Slide 1 text
Scott Sullivan
@scotsullivan
lpk.com
Machine Learning
for Designers
Slide 2
Slide 2 text
No content
Slide 3
Slide 3 text
No content
Slide 4
Slide 4 text
No content
Slide 5
Slide 5 text
WHAT IS MACHINE LEARNING?
Slide 6
Slide 6 text
WHAT IS MACHINE LEARNING?
“improving performance in some task with experience”
-Tom Mitchell, author of Machine Learning
Slide 7
Slide 7 text
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
Slide 8
Slide 8 text
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
Slide 9
Slide 9 text
WHAT IS MACHINE LEARNING?
What’s different?
Slide 10
Slide 10 text
WHAT IS MACHINE LEARNING?
What’s different?
traditional code:
Explicit
(instructions)
Slide 11
Slide 11 text
WHAT IS MACHINE LEARNING?
What’s different?
traditional code:
Explicit
machine learning:
Implicit
(instructions) (train with data)
Slide 12
Slide 12 text
WHAT IS MACHINE LEARNING?
Software that is written by showing it data.
Then it makes predictions.
!
Our Data Model Small part of software
Slide 13
Slide 13 text
WHAT IS MACHINE LEARNING?
Classification
Slide 14
Slide 14 text
WHAT IS MACHINE LEARNING?
Regression
Slide 15
Slide 15 text
WHAT IS MACHINE LEARNING?
Clustering
Slide 16
Slide 16 text
WHAT IS MACHINE LEARNING?
Slide 17
Slide 17 text
PROCESS / SKETCHING
Slide 18
Slide 18 text
PROCESS / SKETCHING
Slide 19
Slide 19 text
PROCESS / SKETCHING
Slide 20
Slide 20 text
MICRO EXERCISE 1:
PSEUDOCODING
Slide 21
Slide 21 text
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
Slide 22
Slide 22 text
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
Slide 23
Slide 23 text
MACHINE LEARNING AS DESIGN
MATERIAL
Slide 24
Slide 24 text
MACHINE LEARNING AS DESIGN MATERIAL
Slide 25
Slide 25 text
MACHINE LEARNING AS DESIGN MATERIAL
Slide 26
Slide 26 text
MACHINE LEARNING AS DESIGN MATERIAL
Slide 27
Slide 27 text
MACHINE LEARNING AS DESIGN MATERIAL
Slide 28
Slide 28 text
MACHINE LEARNING AS DESIGN
ARCHITECTURE
Slide 29
Slide 29 text
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
Slide 30
Slide 30 text
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
Slide 31
Slide 31 text
MACHINE LEARNING AS DESIGN ARCHITECTURE
6
5.5
Slide 32
Slide 32 text
MACHINE LEARNING AS DESIGN ARCHITECTURE
6
5.5
Slide 33
Slide 33 text
MACHINE LEARNING AS DESIGN ARCHITECTURE
6
5.5
Slide 34
Slide 34 text
MACHINE LEARNING AS DESIGN ARCHITECTURE
6
5.5
Slide 35
Slide 35 text
MACHINE LEARNING AS DESIGN ARCHITECTURE
6
5.5
Activity Time went to bed
Alcohol
consumption
Time spent
at bar
Slide 36
Slide 36 text
MACHINE LEARNING AS DESIGN ARCHITECTURE
6
5.5
Activity Time went to bed
Alcohol
consumption
Time spent
at bar
Slide 37
Slide 37 text
MACHINE LEARNING AS DESIGN ARCHITECTURE
6
5.5
Activity Time went to bed
Alcohol
consumption
Time spent
at bar
Slide 38
Slide 38 text
MACHINE LEARNING AS DESIGN ARCHITECTURE
6
5.5
Activity Time went to bed
Alcohol
consumption
Time spent
at bar
Slide 39
Slide 39 text
PROCESS
Slide 40
Slide 40 text
Designers need to do this.
!
Our Data Model Small part of software
Slide 41
Slide 41 text
PROCESS / HANDS ON
Slide 42
Slide 42 text
PROCESS / HANDS ON WITH DATA Tableau
Slide 43
Slide 43 text
PROCESS / SKETCHING
Slide 44
Slide 44 text
PROCESS / SKETCHING
Slide 45
Slide 45 text
PROCESS / SKETCHING
Slide 46
Slide 46 text
PROCESS / IDEATION
Slide 47
Slide 47 text
PROCESS / IDEATION
Slide 48
Slide 48 text
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.
Slide 49
Slide 49 text
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.
Slide 50
Slide 50 text
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?
Slide 51
Slide 51 text
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.
Slide 52
Slide 52 text
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.
Slide 53
Slide 53 text
PROCESS / DIALOG
Slide 54
Slide 54 text
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
Slide 55
Slide 55 text
PROCESS / DIALOG
EXPLICIT FEEDBACK
• Goal setting
• Preferences
• Answering questions
• Manual Adjustments to their
model
USER INPUT
Slide 56
Slide 56 text
PROCESS / DIALOG
Slide 57
Slide 57 text
PROCESS / DIALOG
Slide 58
Slide 58 text
PROCESS / DIALOG
EXPLICIT FEEDBACK
• Goal setting
• Preferences
• Answering questions
• Manual Adjustments to their
model
USER INPUT
IMPLICIT FEEDBACK
• Adjustments to behavior
• Changes in engagement
Slide 59
Slide 59 text
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
Slide 60
Slide 60 text
PROCESS / DIALOG
Slide 61
Slide 61 text
No content
Slide 62
Slide 62 text
DIALOG / PROTOTYPING
Slide 63
Slide 63 text
DIALOG / PROTOTYPING
Slide 64
Slide 64 text
DIALOG / PROTOTYPING
Slide 65
Slide 65 text
DIALOG / ∞
Slide 66
Slide 66 text
DIALOG / ∞
IMPROVEMENT
How would you tell a human how to
improve the predictions?
FEEDBACK
AWARENESS
What assumptions should the model
not make?
Slide 67
Slide 67 text
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?
Slide 68
Slide 68 text
THANK YOU!
Scott Sullivan
LPK
@scotsullivan
LPK.com
Slide 69
Slide 69 text
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.