Cognitive Biases in Engineering Organizations

Velocity Barcelona – November 17, 2014 @jonathanklein Jonathan Klein in
Engineering Organizations Cognitive Biases

Talk Resources (Slides/Links) http://jkle.in/biases

jkle.in/biases @jonathanklein 3

jkle.in/biases @jonathanklein Who Am I? 3

jkle.in/biases @jonathanklein Who Am I? • Senior performance engineer at
Etsy 3

Etsy • I write the Etsy Site Performance Reports 3

Etsy • I write the Etsy Site Performance Reports • I like figuring out why we believe what we believe 3

@jonathanklein jkle.in/biases Cognitive Bias: A deviation in judgement where inferences
may be illogical 5

@jonathanklein jkle.in/biases System 1 Operates automatically and quickly, with little
or no effort and no sense of voluntary control 6

@jonathanklein jkle.in/biases System 2 Allocates attention to the effortful mental
activities that demand it, including complex computations 7

@jonathanklein jkle.in/biases System 1 Operates automatically and quickly, with little
or no effort and no sense of voluntary control 8

@jonathanklein jkle.in/biases 5-10% of videos were upside-down 11

“It was designed for right-handed users, but phones are usually
rotated 180 degrees when held in left hands.” 12

jkle.in/biases @jonathanklein • Projection Bias • Planning Fallacy • Bandwagon
Effect • Sunk Cost Fallacy • Hyperbolic Discounting • Fundamental Attribution Error 13

Part 1: A Fictional Story

@jonathanklein jkle.in/biases Sales Guy 16

@jonathanklein jkle.in/biases Engineering Manager 17

@jonathanklein jkle.in/biases Software Engineer 18

@jonathanklein jkle.in/biases CEO 19

@jonathanklein jkle.in/biases Sales Guy Sells a feature that doesn’t exist
yet 21

@jonathanklein jkle.in/biases Engineering Manager ROI is positive at < 6
weeks, thinks we can do it in 4 22

@jonathanklein jkle.in/biases Software Engineer “Yeah, we can do that” 23

@jonathanklein jkle.in/biases Two weeks of development… 24

@jonathanklein jkle.in/biases Software Engineer “Oh sh*t, we need 8 more
weeks” 25

@jonathanklein jkle.in/biases Engineering Manager “Let’s push forward, and hope we
beat expectations” 26

@jonathanklein jkle.in/biases Software Engineer “Okay, we’ll cut some corners” 27

@jonathanklein jkle.in/biases Six more weeks go by… 28

@jonathanklein jkle.in/biases CEO “Who do we need to fire?” 30

This could have been avoided 31

Deconstructing the Story

@jonathanklein jkle.in/biases Sales Guy Sells a feature that doesn’t exist
yet 33

@jonathanklein jkle.in/biases Projection Bias One thinks that others have the
same priority, attitude or belief that one harbors oneself, even if this is unlikely to be the case. 34

same priority, attitude or belief that one harbors oneself, even if this is unlikely to be the case. Solution 34

same priority, attitude or belief that one harbors oneself, even if this is unlikely to be the case. Solution Empathy 34

same priority, attitude or belief that one harbors oneself, even if this is unlikely to be the case. Solution Empathy Have people switch roles 34

@jonathanklein jkle.in/biases Planning Fallacy The tendency for people to underestimate
how long they will need to complete a task, even when they have experience of similar tasks over-running 36

how long they will need to complete a task, even when they have experience of similar tasks over-running Solution 36

how long they will need to complete a task, even when they have experience of similar tasks over-running Solution Prototyping 36

how long they will need to complete a task, even when they have experience of similar tasks over-running Solution Prototyping Use information from similar ventures 36

how long they will need to complete a task, even when they have experience of similar tasks over-running Solution Prototyping Use information from similar ventures “Planning poker” exercise 36

@jonathanklein jkle.in/biases Planning Poker Cards 37

@jonathanklein jkle.in/biases 38

@jonathanklein jkle.in/biases Software Engineer “Yeah, we can do that” 39

@jonathanklein jkle.in/biases Bandwagon Effect The rate of uptake of beliefs,
ideas, fads and trends increases the more that they have already been adopted by others 40

ideas, fads and trends increases the more that they have already been adopted by others Solution 40

ideas, fads and trends increases the more that they have already been adopted by others Solution Make decisions asynchronously 40

ideas, fads and trends increases the more that they have already been adopted by others Solution Make decisions asynchronously Ask for confidential feedback 40

ideas, fads and trends increases the more that they have already been adopted by others Solution Make decisions asynchronously Ask for confidential feedback Don’t prime the estimate 40

ideas, fads and trends increases the more that they have already been adopted by others Solution Make decisions asynchronously Ask for confidential feedback Don’t prime the estimate “Planning poker” exercise 40

@jonathanklein jkle.in/biases Software Engineer “Oh sh*t, we need 8 more
weeks” 41

@jonathanklein jkle.in/biases Engineering Manager “Let’s push forward, and hope we
beat expectations” 42

@jonathanklein jkle.in/biases Sunk Cost Fallacy A sunk cost is a
retrospective (past) cost that has already been incurred and cannot be recovered 43

retrospective (past) cost that has already been incurred and cannot be recovered Solution 43

retrospective (past) cost that has already been incurred and cannot be recovered Solution Do some math 43

retrospective (past) cost that has already been incurred and cannot be recovered Solution Do some math At every point in the project, the ROI should still be positive 43

@jonathanklein jkle.in/biases 44 Return on Investment

@jonathanklein jkle.in/biases 45 Return on Investment

@jonathanklein jkle.in/biases 45 Return on Investment Project Completed

@jonathanklein jkle.in/biases Our Example Estimated Development Time: 4 Weeks Estimated
Development Cost: $8000 Client Is Paying: $12,000 46

@jonathanklein jkle.in/biases Our Example Estimated Development Time: 8 Weeks Estimated
Development Cost: $16,000 Client Is Paying: $12,000 47

@jonathanklein jkle.in/biases Software Engineer “Okay, we’ll cut some corners” 49

@jonathanklein jkle.in/biases Hyperbolic Discounting Given two similar rewards, humans show
a preference for one that arrives sooner rather than later. Humans are said to discount the value of the later reward, by a factor that increases with the length of the delay. 50

a preference for one that arrives sooner rather than later. Humans are said to discount the value of the later reward, by a factor that increases with the length of the delay. Solution 50

a preference for one that arrives sooner rather than later. Humans are said to discount the value of the later reward, by a factor that increases with the length of the delay. Solution Deliberately stop and do an estimation exercise 50

a preference for one that arrives sooner rather than later. Humans are said to discount the value of the later reward, by a factor that increases with the length of the delay. Solution Deliberately stop and do an estimation exercise Consider implications for future developers 50

@jonathanklein jkle.in/biases The Stanford Marshmallow Experiment 51

jkle.in/biases @jonathanklein Experimental Outcomes Researchers found that children who were
able to wait longer for the preferred rewards tended to have better life outcomes: • SAT Scores • Educational attainment • Body Mass Index (BMI) 55

@jonathanklein jkle.in/biases CEO “Who do we need to fire?” 57

@jonathanklein jkle.in/biases Fundamental Attribution Error People's tendency to place an
undue emphasis on internal characteristics to explain someone else's behavior in a given situation, rather than considering external factors. 58

@jonathanklein jkle.in/biases Fundamental Attribution Error People's tendency to place an
undue emphasis on internal characteristics to explain someone else's behavior in a given situation, rather than considering external factors. Solution Blameless postmortems 59

@jonathanklein jkle.in/biases Everyone had good intentions 60

Our intuition is imperfect 61

Part 2: Real Studies

@jonathanklein jkle.in/biases Which is More Dangerous? 64

@jonathanklein jkle.in/biases Which is More Dangerous? A disease that kills
1,286 people out of every 10,000 64

@jonathanklein jkle.in/biases Which is More Dangerous? A disease that kills
1,286 people out of every 10,000 A disease that kills 24.4 out of 100 64

@jonathanklein jkle.in/biases Which is More Dangerous? ! A disease that
kills 1,286 people out of every 10,000 A disease that kills 24.4 out of 100 65

kills 1,286 people out of every 10,000 A disease that kills 24.4 out of 100 65 Most people say this

kills 1,286 people out of every 10,000 A disease that kills 24.4 out of 100 65 Most people say this This is twice as bad

@jonathanklein jkle.in/biases Denominator Neglect 66

@jonathanklein jkle.in/biases We Can’t Always Trust Our Intuition 67

@jonathanklein jkle.in/biases Dinner Plates Soup/Salad bowls Dessert plates Cups Saucers
8, all good condition 8, all good condition 8, all good condition 8, 2 broken 8, 7 broken 8, all good condition 8, all good condition 8, all good condition None None 68 Set A: 40 Pieces Set B: 24 Pieces

jkle.in/biases @jonathanklein Three Groups in the Experiment ! 1. Shown
both sets together 2. Shown only set A 3. Shown only set B 70

@jonathanklein jkle.in/biases Group 1 - Joint Evaluation: Set A: $32
Set B: $30 71

@jonathanklein jkle.in/biases Groups 2 and 3 - Single Evaluation: 72

@jonathanklein jkle.in/biases Groups 2 and 3 - Single Evaluation: Set
A: $23 Set B: $33 72

Manipulation

jkle.in/biases @jonathanklein 75

jkle.in/biases @jonathanklein Subscriptions to The Economist 75

jkle.in/biases @jonathanklein Subscriptions to The Economist 1. A web-only subscription
for $59 75

for $59 2. A print-only subscription for $125 75

for $59 2. A print-only subscription for $125 3. A web + print subscription for $125 75

jkle.in/biases @jonathanklein Subscriptions to The Economist ! 1. A web-only
subscription for $59 2. A print-only subscription for $125 3. A web + print subscription for $125 76 Useless

subscription for $59 2. A print-only subscription for $125 3. A web + print subscription for $125 77 16% 0% 84%

subscription for $59 2. A print-only subscription for $125 3. A web + print subscription for $125 78

subscription for $59 2. A print-only subscription for $125 3. A web + print subscription for $125 79 16% 84% 68% 32%

@jonathanklein jkle.in/biases Removing the middle option costs $3400+ per 100
subscribers 80

@jonathanklein jkle.in/biases Based on 2013 circulation, this could cost The
Economist $53M 81

@jonathanklein jkle.in/biases What’s the best way to sell a $3,000
suit? 82

@jonathanklein jkle.in/biases What’s the best way to sell a $3,000
suit? Put it next to a $10,000 suit 82

@jonathanklein jkle.in/biases Revisiting the Sunk Cost Fallacy 87

@jonathanklein jkle.in/biases It Preys on Loss Aversion 88

@jonathanklein jkle.in/biases Another Experiment ! 89 $0.01 $0.15

@jonathanklein jkle.in/biases Another Experiment ! 90 Most people choose this
$0.01 $0.15

@jonathanklein jkle.in/biases Lower the Price ! 91 Free! $0.14

@jonathanklein jkle.in/biases Lower the Price ! 92 Most people choose
this Free! $0.14

@jonathanklein jkle.in/biases We need a framework for decision making 94

Bayesian Reasoning

@jonathanklein jkle.in/biases 96 Bayes' Theorem

@jonathanklein jkle.in/biases 1. Anchor your judgement of the probability of
an outcome on a plausible base rate 97

@jonathanklein jkle.in/biases 2. Question the importance and relevancy of your
evidence 98

@jonathanklein jkle.in/biases “What is the base rate for estimation accuracy?”
100

@jonathanklein jkle.in/biases “Why is this situation different?” 101

@jonathanklein jkle.in/biases To make rational decisions, you must understand the
base rate 107

@jonathanklein jkle.in/biases You must also objectively evaluate the evidence for
departing from the base rate 108

Two Principles: Dramatically better decisions 109

@jonathanklein jkle.in/biases 110 jkle.in/biases

@jonathanklein jkle.in/biases Thanks! jkle.in/biases @jonathanklein [email protected] 111

Cognitive Biases in Engineering Organizations

Cognitive Biases in Engineering Organizations

More Decks by Jonathan Klein

Other Decks in Science

Featured

Transcript