Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Augmenting Human Decision Making with Data Science - PyCon
Search
Kelsey Pedersen
May 11, 2018
Technology
0
99
Augmenting Human Decision Making with Data Science - PyCon
Kelsey Pedersen
May 11, 2018
Tweet
Share
More Decks by Kelsey Pedersen
See All by Kelsey Pedersen
Defying the Odds: How to Apply (and Get Selected!) to Speak at Conferences
kelseypedersen
0
37
Augmenting Human Decision Making with Data Science - RubyxElixir Conf 2018
kelseypedersen
1
90
RubyConf_2017_-_Kelsey_Pedersen.pdf
kelseypedersen
0
28
RubyConf_2017_-_Kelsey_Pedersen.pdf
kelseypedersen
0
19
Augmenting Human Decision Making with Data Science
kelseypedersen
1
85
Other Decks in Technology
See All in Technology
オーナーシップを持つ領域を明確にする
konifar
13
3.2k
チームでロジカルシンキングに改めて向き合っている話 〜学習環境と実践⽅法〜
sansantech
PRO
3
2.5k
生成AIの変革の時代に、直近1年で直面した課題とその解決策
ktc_wada
0
290
KubeCon EU 2024 Recap “Kubernetes Policy Time Machine: Where to Next?”
ryysud
0
220
Azureの基本的な権限管理の勉強会
yhana
0
560
KubeConにproposalを送りたい人へのアドバイス
sat
PRO
3
250
require(ESM)とECMAScript仕様
uhyo
3
720
VS CodeでAWSを操作しよう
smt7174
8
1.7k
今年のRubyKaigiはProfiler Year🤘
osyoyu
0
120
On Your Data を超えていく!
hirotomotaguchi
2
690
ChatGPT for IT Service Management (IT Pro)
dahatake
7
1.6k
競技としてのKaggle、役に立つKaggle
yu4u
3
1.7k
Featured
See All Featured
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
20
1.9k
Building Effective Engineering Teams - LeadDev
addyosmani
28
1.8k
Stop Working from a Prison Cell
hatefulcrawdad
266
19k
GraphQLの誤解/rethinking-graphql
sonatard
50
9.2k
Facilitating Awesome Meetings
lara
42
5.6k
A Philosophy of Restraint
colly
197
16k
Design by the Numbers
sachag
274
18k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
155
14k
Teambox: Starting and Learning
jrom
128
8.4k
10 Git Anti Patterns You Should be Aware of
lemiorhan
648
58k
Typedesign – Prime Four
hannesfritz
36
2.1k
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
125
32k
Transcript
HUMAN DECISION MAKING Kelsey Pedersen Software Engineer, Stitch Fix with
DATA SCIENCE AUGMENTING @kelsey_pedersen
72% of Americans are scared of computers taking over our
jobs @kelsey_pedersen
51% worry about gun controls
55% worry about affordable health care
Gun Controls Health Care Algorithms & Robots 51% 55% 72%
@kelsey_pedersen
None
@kelsey_pedersen
How to balance human decisions with algorithmic decisions in our
software GOAL
@kelsey_pedersen KELSEY PEDERSEN
@kelsey_pedersen
None
What are the limitations of human decision making? What are
the limitations of algorithmic decision making? How can human decisions be augmented by data science? @kelsey_pedersen
What are the limitations of human decision making?
System 1 System 2
Intuition & feelings System 1
FAST AUTOMATIC UNCONSCIOUS System 1 @kelsey_pedersen
Interpretation of our surroundings System 1
None
None
None
None
95% of human decisions are made in System 1
Analytical & effortful System 2
None
1737 x 1990 Answer: 3,456,630 @kelsey_pedersen
None
System 1 gut feelings System 2 computations
@kelsey_pedersen
Gut feelings are unpredictable @kelsey_pedersen
Environment & mood Influence thoughts and feelings
INCONSISTENT INFO OVERLOAD BIASED
Gut feelings are driven by your own views and preferences
@kelsey_pedersen
Biases occur outside of our own awareness Cause us to
think and act irrationally
Computation of lots of info takes time and energy @kelsey_pedersen
Lots of information Causes physical response
INCONSISTENT || BIASED || INFO OVERLOAD
Our stylists are human, so these inconsistencies apply to them
too. @kelsey_pedersen
INCONSISTENT INFO OVERLOAD BIASED
Inconsistent Judgments by Stylists
Biased Decisions by Stylists
Information Overload by Stylists
@kelsey_pedersen
Case Study @kelsey_pedersen
Our data science team uses multiple sources of internal data
@kelsey_pedersen Direct from our customers
Did a customer keep or return an item? Customer survey
* What size top are you? S, M, L, XL Buying Patterns Size of this item? Too small, just right, too big Checkout feedback * Helps the cold start problem Direct Feedback Indirect Feedback
FIT @kelsey_pedersen
FIT STYLE @kelsey_pedersen
FIT STYLE PRICE @kelsey_pedersen
FIT STYLE PRICE SIZE @kelsey_pedersen
@kelsey_pedersen
How can human decisions be augmented by data science?
@kelsey_pedersen
(1) Stylist deciding on one item
None
72% 60% 48% 44% @kelsey_pedersen
@kelsey_pedersen
@kelsey_pedersen
@kelsey_pedersen
(2) Stylist deciding on all 5 items
72% 60% 48% 44% 56% @kelsey_pedersen
50% 26% 18% 14% 27% @kelsey_pedersen
@kelsey_pedersen
@kelsey_pedersen
(3) One Client’s Feedback on all 5 items
@kelsey_pedersen
@kelsey_pedersen
None
Stylists use feedback to train to make more accurate decisions
in the future @kelsey_pedersen
(4) All Clients’ Feedback for one Stylist
STATS
None
Stylist Stylist Manager
(5) All Clients’ Feedback overtime for all Stylists
None
None
Deciding on one item Client Feedback on all 5 items
Deciding on all 5 items @kelsey_pedersen Client Feedback overtime for one Stylist All Feedback overtime for all Stylists GUIDE TRAIN
GUIDE DECISIONS WITH COMPUTATIONS TRAIN DECISIONS WITH FEEDBACK
What are the limitations of algorithmic decision making?
None
None
None
Stylists are able to override the algorithms. @kelsey_pedersen
None
72% 60% 48% 10% @kelsey_pedersen
72% 60% 48% 44% 24% @kelsey_pedersen
When intuition doesn’t match the algorithm, we can learn from
that. @kelsey_pedersen
None
None
stylists data scientists @kelsey_pedersen
What is the future of humans and data science?
@kelsey_pedersen
None
System 1 gut feelings System 2 computations &
@kelsey_pedersen
Data Science Humans System 1 System 2 Data Science Humans
System 3 @kelsey_pedersen
Predictive & intuitive System 3
None
System 1 gut feelings System 2 computations System 3 feedback
What is the business impact of system 3?
lower labor cost fewer mistakes from humans greater client satisfaction
increased keep rate @kelsey_pedersen
In conclusion…
Conclusion 95%
INCONSISTENT INFO OVERLOAD BIASED
None
Conclusion @kelsey_pedersen
Humans lack the ability to process large volumes of information.
@kelsey_pedersen
Machines lack intuition, empathy, nuance and ethics. @kelsey_pedersen
None
@kelsey_pedersen
Conclusion @kelsey_pedersen
humans data science @kelsey_pedersen
Thank you!
@kelsey_pedersen Thanks! KELSEY PEDERSEN Stitch Fix Promo bit.ly/pycon-stitchfix We’re hiring!