This document contains a reading list and notes from a psychology for startups lecture. The key points are:
1) Mental models define how we think the world works and influence all of our decisions, but they are not always accurate. We must actively seek to improve our models.
2) Our minds are prone to predictable errors and biases like availability bias, representativeness bias, and anchoring. We must be aware of these to avoid irrational decisions.
3) When seeking advice, focus on those with broad but not too deep expertise ("foxes"), rather than experts in a single area ("hedgehogs"). Foxes are more likely to give good long-term advice.
4)
1. Justin Singer - justin.e.singer@gmail.com http://msnbcmedia.msn.com/i/MSNBC/Components/Photo/_new/Afghanistan_Dynamic_Planning.pdf
Psychology for Startups 19 February 2013
2. Reading list: http://bit.ly/WVxDCS
• Psychology of Intelligence Analysis: http://1.usa.gov/12K7Wc1
- Chapter 1 - Thinking about Thinking
- Chapter 2 - Perception
- Chapter 4 - Strategies for Analytical Judgment
- Chapter 6 - Keeping an Open Mind
• Everybody’s an Expert: http://nyr.kr/WVwviv
• Munger’s Worldly Wisdom: http://bit.ly/WVwxXQ
• Wikipedia’s List of cognitive biases: http://bit.ly/1332wsr
• David Foster Wallace - This is Water
- Part 1: http://bit.ly/W2D4RM
- Part 2: http://bit.ly/W2DgR8
• The Psychology of Human Misjudgment: http://bit.ly/15tDl1N
• The Design of Everyday Things: http://amzn.to/12KctuP
3. Why psychology?
Product
Strategy
Hiring
Managing
Marketing
Entrepreneurship depends on robust models of learning
habit
behavior
desire
interaction
expectation
4. Today’s arguments
• Pay close attention to mental models -- they’re the
basis for everything
• Our minds are broken, but in predictable ways
• The most important choice you will make is whose
advice to take
• Fuck it. Keep moving forward
6. What are mental models?
“[M]odels people have of themselves, others, the
environment, and the things with which they interact."
- Donald A. Norman. The Design of Everyday Things (1988)
10. What are mental models?
Mental models define how we think the world works,
but not necessarily how it actually works
- Me, just now
Mental models are necessarily personal
If a model doesn’t work for you, build a better one
When judging a model’s quality, focus on process, not outcome
11. How do we form mental models?
Real world What a video camera would record.
Interpretation The story we create in our mind.
Is our story confirmed or disconfirmed?
Feedback (usually we only ask the former)
12. Single-loop learning
Real world
Information
Decision feedback
Decision making Mental
rules model
http://en.wikipedia.org/wiki/Mental_model
13. Single-loop learning
“Insanity is repeating the same mistakes and expecting
different results.”
- Narcotics Anonymous. Basic Text, pg. 11
(nope, not Einstein)
Want better results? Change your model
http://amonymifoundation.org/uploads/NA_Approval_Form_Scan.pdf
14. Double-loop learning
Real world
Information
Decision feedback
Decision making Mental
rules model
http://en.wikipedia.org/wiki/Mental_model
15. Learning loops in Product Design
What’s missing?
Donald A. Norman. The Design of Everyday Things (1988).
16. Learning loops in Product Design
User feedback should alter the
product by altering the design model
Donald A. Norman. The Design of Everyday Things (1988).
17. And remember...
Just because people are using the same words,
doesn’t mean they are thinking the same thing
http://guide.cred.columbia.edu/guide/sec1.html
18. Strong sources of mental models
• Physical laws (especially movement mechanics)
- Elasticity (springs)
- Friction
• Large and representative data sets (empirical
observation)
• Careful experimentation (seeking to disconfirm)
• Relevant analogy
19. Weak sources of mental models
• Abstract theory
• Personal experience
• Irrelevant analogy
• Repeated observations (small data sets)
• Single observation (single data point)
• Anecdote/inductive reasoning (Malcolm Gladwell)
• Opinion
Unfortunately, the less data we have,
the more heavily we weight it
21. What are heuristics?
Heuristics are simple, efficient rules people use to
form judgments and make decisions
Heuristics usually work well, but can lead to systematically
irrational outcomes. These errors are called biases
Key people to know: Herbert A. Simon, Amos Tversky, Daniel Kahneman
22. Three major heuristics to know
Overweights the probability of events
Availability that are recent, vivid, or dramatic
Overweights the probability of events
Representativeness
that match our expectations
Anchoring and Overweights the importance of the first
adjustment piece of information we receive
23. Availability heuristic
The more vivid or recent an event, the more likely
we are to overestimate its likelihood
24. Availability heuristic
Deaths vs. Dollars
Annual deaths Annual spending ($B)
597,689 Heart Disease $2.049
574,743 Cancer $5.448
69,071 Diabetes $1.076
83,494 Alzheimer’s $0.448
35,332 Car Accidents $0.867
NHTSA budget
All deaths since 2000 3,023 Terrorism $6.814
TSA budget
http://en.wikipedia.org/wiki/Transportation_Security_Administration
http://report.nih.gov/categorical_spending.aspx
http://www.dot.gov/mission/budget/nhtsa-fy-2010-budget-estimate
http://www.state.gov/j/ct/rls/crt/
http://www.cdc.gov/nchs/fastats/deaths.htm
25. Availability heuristic
How feature creep happens
Just because a few people bitch about it doesn’t mean
you should change it. Dig deeper and use your judgment
https://twitter.com/vacanti/status/184003264361148416
26. Representativeness heuristic
The fact that something “looks” like you’d expect
does not make it more likely to be what you’re
looking for
28. Representativeness heuristic
What does random look like?
Random
HHHHHTTTTH
HTHHHTHTHT
Not random
Gambler’s fallacy: the belief that small samples will
reflect the populations they’re drawn from
29. Proof by example
We tend to vastly overweight the evidentiary value of
small, not necessarily representative samples
30. Base rate fallacy
When making judgments, we tend to ignore prior
probabilities and focus on expected similarities
http://www.businessinsider.com/how-andreessen-horowitz-chooses-investments-2013-2?op=1
To be fair, this is a bit of a cherry pick -- the next slide in the deck is more nuanced
31. Representativeness heuristic :: hiring
What does a designer look like?
http://karakreative.blogspot.com/2013/02/graphic-designer-of-month-paul-rand.html http://tech.fortune.cnn.com/2011/06/27/quoras-designing-woman/
http://vimeo.com/putorti http://topics.nytimes.com/top/reference/timestopics/people/f/shepard_fairey/index.html
32. Representativeness heuristic :: hiring
Designers look like everyone else!
Paul Rand Rebekah Cox
Jason Purtorti
Shepherd Fairey
http://karakreative.blogspot.com/2013/02/graphic-designer-of-month-paul-rand.html http://tech.fortune.cnn.com/2011/06/27/quoras-designing-woman/
http://vimeo.com/putorti http://topics.nytimes.com/top/reference/timestopics/people/f/shepard_fairey/index.html
33. Representativeness heuristic :: hiring
Who do you want to work with?
• Great people are... • Great people are not necessarily...
- Thoughtful - Ex-FB/Paypal/Google/etc.
(also, fundamental attribution error)
- Productive
- Team-oriented
- Graduates of Stanford/CMU/
Wharton/Columbia/college
- Quick studies
- Arrogant
- Patient teachers
- Overly deferential
- Empathetic
- Aggressively passionate
- Pragmatic
- On Twitter
- Comfortable with
- Morally superior
uncertainty
- A strong cultural fit
- “Design-y”
34. Representativeness heuristic :: skill vs. luck
Fundamental attribution error
We tend to overvalue personality-based explanations and
undervalue situational explanations for the actions of others
Self-serving bias
We tend to attribute our successes to personal/internal factors
and attribute our failures to situational/external factors
35. What’s more likely?
That a large group of Super Businessmen happened to work
together at Paypal...
Or, that a large group of smart people happened to meet and
work together at the right place at the right time?
http://money.cnn.com/2007/11/13/magazines/fortune/paypal_mafia.fortune/index.htm
36. What’s more likely?
That a large group of Super Businessmen happened to work
together at Fairchild Semiconductor...
Or, that a large group of smart people happened to meet and
work together at the right place at the right time?
http://www.inc.com/articles/201109/then-and-now-venture-capital.html
37. Representativeness heuristic :: skill vs. luck
Judging outliers
When it comes to judging outliers, we tend to overestimate
the effect of skill and wildly underestimate the effect of luck
The law of exponential returns
Any great entrepreneur can build a $10M* business on skill
No great entrepreneur can build a $1B business without luck
* Amounts aren’t meant to be taken literally
38. Anchoring and adjustment
The tendency to base subsequent judgments on the
first piece of information we gather (even when the
information is entirely irrelevant)
39. Anchoring and adjustment
Negotiating strategies
• When you receive a lowball offer, reject it out of hand
(i.e., don’t make a counteroffer)
• Corollary: if making the first offer, aim for just beyond acceptable
(i.e., not so high or low as to elicit rejection)
• Don’t send an agreeable person to the negotiating table
• Decide walkaway points before negotiating and stick to them
• Be wary of framing effects
• Smile! Sadness tends to exacerbate the anchoring effect
• Practice! Anchoring effects diminish with experience
40. “The fox knows many things;
the hedgehog one great thing.”
- Archilochus
Expert Prediction
http://www.etsy.com/listing/60007735/woodland-animal-pair-hedgehog-and-fox
41. What does this have to do with startups?
Every feature
suggestion
opinion
piece of advice is a prediction
Who should you listen to?
How much credence should you give?
42. What will Facebook
close at on its IPO day?
http://collider.com/mark-zuckerberg-reviews-the-social-network/
http://www.theatlantic.com/technology/archive/2012/05/twitter-tech-elite-seriously-overstimated-facebooks-closing-price/257406/
43. Oopsies...
$38*
* required significant price support from underwriters
http://collider.com/mark-zuckerberg-reviews-the-social-network/
http://www.theatlantic.com/technology/archive/2012/05/twitter-tech-elite-seriously-overstimated-facebooks-closing-price/257406/
46. "Freddie Mac and Fannie Mae are fundamentally sound. They're not in
danger of going under…I think they are in good shape going
forward."
- Barney Frank (D-Mass.) House Fin. Svcs. Comm. chairman, July 14, 2008
Placed into conservatorship in September
"I think you'll see [oil prices at] $150 a barrel by the end of the year"
- T. Boone Pickens, May 20, 2008
$100/bbl in May - $135/bbl in July - $38/bbl in November
“The subscription model of buying music is bankrupt. I think you
could make available the Second Coming in a subscription model and
it might not be successful.”
- Steve Jobs, Rolling Stone, Dec. 3, 2003
Spotify and Rdio would beg to differ
47. These are very, very smart people who
were very, very wrong.
Why?
48. What does it mean to be T-shaped?
http://www.stratabridge.com/2011/08/putting-the-t-into-leadership/t-shaped/
49. One model for thinking about advisors
Fox Fox-Experts Hedgehog-Experts
Knows many things well
Hedgehog
Knows one thing well
Expert
Expert in the subject at hand Fox-Dilettantes Hedgehog-Dilettantes
Dilettante
Expert in a related subject
(but not the one at hand)
When it comes to China, the Chinese Ambassador is an expert
and the British Ambassador is a dilettante
50. Refers to political extremism
regardless of party
Tetlock, Philip E., Expert Political Judgment: How Good Is It? How Can We Know? (2005), fig. 3.4
51. If advice is a prediction, then whose advice
deserves your attention?
Short-term advice Long-term advice
1. Fox-Experts 1. Fox-Dilettantes
2. Fox-Dilettantes 2. Fox-Experts
3. Hedgehog-Dilettantes 3. Hedgehog-Dilettantes
4. Hedgehog-Experts 4. Hedgehog-Experts
Turns out that a lot of knowledge in
a single area is a dangerous thing
52. How to recognize a fox
• skeptical of deductive approaches to explanation and prediction
• disposed to qualify tempting analogies by noting disconfirming
evidence
• reluctant to make extreme predictions of the sort that start to
flow when positive feedback loops go unchecked by dampening
mechanisms
• worried about hindsight bias causing us to judge those in the
past too harshly
• prone to a detached, ironic view of life
• motivated to weave together conflicting arguments on
foundational issues in the study of politics, such as the role of
human agency or the rationality of decision making
Tetlock, Philip E. Expert Political Judgment: How Good Is It? How Can We Know? 2006.
53. There is no textbook for this
“Everyone is totally blind, feeling around in the dark, trying
to succeed at building this thing we call a ‘business’.”
- Dan Shipper
The best you can hope for is to develop
a robust learning process
http://danshipper.com/how-to-make-a-million-dollars
54. Treat your models as hypotheses
Make sure they’re testable
Models that can’t be disproven are aren’t model -- they’re beliefs
Actively seek to disprove them
Welcome disproof -- a model disproved is a lesson learned
Look for hidden assumptions
Treat secondhand data as assumptions until proven otherwise
Question their predictability
The same event may be evidence of many different hypotheses
Models don’t care about your loyalty
If a model doesn’t work, change it
55. Uncertainty stops most people in their tracks, but it’s
only by movement that uncertainty can be resolved
In the meantime, read widely
think deeply
stay humble
chose your advisors wisely
improve your model set
move forward.
“Strong opinions, weakly held.”
- Paul Saffo
56. “Our brains have just
one scale, and we
resize our experiences
to fit.”
http://xkcd.com/915/
57. Jerry Neumann - mti@neuvc.com
Justin Singer - justin.e.singer@gmail.com