Notes on Thinking Fast and Slow by Daniel Kahneman

I just finished reading Thinking, Fast and Slow by Daniel Kahneman, and I absolutely loved it. I think you can tell based on the sheer quantity of notes I took, and I try to only take notes on things that are new or noteworthy for me personally (and I've read a lot of psychology and behavioral economics literature before).

I think this book is the most comprehensive and clear coverage of the field I've seen to date. No wonder so many people reference it and recommend it.

I found a lot of similarity in this book to Antifragile and some of Taleb's other books (and Kahneman references Taleb several times in this one).

So many good ideas in here, so many interesting experiments, and such a thorough treatment of cognitive biases, rational decision-making under uncertainty, the remembering vs. experiencing self (and how memories can be manipulated), and so many other interesting topics.​

Intro

• Expectations of intelligent gossip about us by others is motivation
• Systematic errors

Origins

• Studying intuitive statistics with tversky
• Skill and heuristics both sources for intuition
• Intuition is about recognition
• When question is difficult, we usually just answer the easier question and don't notice the substituting
• System 1 fast and 2 slow

Part 1: 2 systems

Ch 1: the characters of the story

• System 1 produces the beliefs for system 2
• 2 needs constant attention and focus to work
• Focusing attention can make people blind
• We are also blind to our blindness
• 2 turns on when 1 is surprised
• 1 has biases
• 1 cannot be turned off
• 2 is in charge of self control
• Cannot prevent system 1 for falling for illusion even if u know different
• Cognitive illusions also
• Easier to see mistakes in others than yourself
• System 1 and 2 fictitious concepts

2 attention and effort

• 2 is lazy
• Mental effort
• Pupils dilate with mental effort
• Pupils as index for amount of mental effort in real time
• Law of least effort: people do whatever is minimal effort
• 2 only one that can deal with multiple things at once and conscious combination of rules
• Most effortful form of slow thinking is when u have to think fast

3 the lazy controller

• Self control is also mental work
• 1 has more influence on behavior when 2 is busy
• Self control requires effort
• 2 runs self control of thoughts and behavior
• Ego depletion
• Flow
• Restoring glucose (real not Splenda) to brain restores ego depletion
• Lazy system 2
• 2 checks if 1 is making a mistake
• Many ppl overconfident in intuitions and avoid mental effort too much
• Cognitive control linked to intelligence

4 associative machine

• Associative activation from simple unconnected words
• Body reacts to words as if reality
• U think with your body
• Ideometer effect: priming affects physical action
• Reciprocal priming: moving quickly makes you less likely to think of old age; forcing smile or frown by accident affects emotions

5 cognitive ease

• Illusion of familiarity when looks clearer and more familiar things look clearer
• Illusion of truth: familiar answers seem true
• Things that make brain run easily feel true
• Repetition creates truth
• Increase legibility
• Increase contrast of text
• Make message simpler
• Put message in rhyming verse
• Quote people with names easy to pronounce
• Creativity as strong associative memory
• Happy subjects better at intuition and system 1

6 norms, surprises, and causes

• Normality of events affected by memory of prior similar events
• Pattern affects view of normality
• Shared cultural norms about objects and world in system 1
• Automatic search for causality
• Need for coherence and narrative

7 machine for jumping to conclusions

• When uncertain, system 1 bets on an answer
• Unbelieving is a function of system 2
• When system 2 is busy (and it is always lazy), we will always believe anything
• Confirmation bias: people seek data to confirm beliefs not refute them
• Halo effect: if like someone then like things from them
• Evidence from first impression lingers and snowballs
• Sequence of getting information matters; latter info discounted
• Decorrelated error: wisdom of crowd in evaluating same data but must be independent and see same info separately
• Before group discussion ask individuals to review separately and write down own conclusions
• What you see is all there is: jumping to conclusions based only on info u see but ignoring info u don't have yet
• Overconfidence
• Framing effects
• same info presented differently understood differently
• Base rate neglect: specific details more memorable than base statistical info unstated

8 how judgments happen

• Basic assessments from system 1
• How things going
• Threats
• Anything special
• Escape or approach
• Shape of face tells dominance
• Facial expression tells trustworthiness
• Ability to assess strangers instantly
• Voters follow facial features seen as competent within 1 second glance (explain 70% of election results)
• Sets and prototypes
• System 1 easily does average visually and also count of objects when around 4-5
• System 2 needed to do summation
• Intensity matching: 1 can do quickly
• Mental shotgun: we often compute more than we need
• 1 does more than 2 needs it to
• Prompt to do one thing often leads automatically to doing or thing something else

9 answering an easier question

• Ppl rarely stumped
• Have opinions about everything even when don't understand
• Substituting questions
• Target question too hard but heuristic question is similar but which u do know
• Heuristic questions usually about your feelings
• 3D heuristic: 3D interpretation of flat printed images; true illusions
• Substitution of 3 dimensions for 2
• Mood heuristic for happiness
• Happiness difficult thing to evaluate in self
• Present state of mind affects Current view of ur happiness
• Affect heuristic: liking something affects how u see it
• Personality of system 2

Part 2 heuristics and biases

10 law of small numbers

• 1 very quick at identifying causes and patterns even when sample size too small
• General statistics hard for 1 to integrate
• Small samples yield extreme results much more likely than larger samples
• Simple statistical fact, no causal connection to be inferred; nothing to explain
• Accident of sampling
• Artifacts: observations purely due to method of sampling and size
• Law of small numbers: experts pick samples too small for research
• Statistical intuitions very poor
• Bias of confidence over doubt
• Ppl are not adequately sensitive to sample size
• We pay more attention to story than source
• 1 cannot distinguish levels of belief
• Only 2 is capable of doubt
• Associative machinery seeks causes
• Random processes produce sequences that convince people that the processes are not random but the people are wrong
• Small schools not better on average; just more likely to produce extreme outcomes

11 anchors

• When people consider a particular value for an unknown quantity before estimating that quantity
• Anchoring as adjustment process
• Insufficient adjustment
• Adjustment from reasons to move away from anchor
• When ppl tired they adjust less from anchor; adjustment is function of system 2
• Anchoring is case of suggestion
• Suggestion is a priming effect through associative machinery that creates errors

12 the science of availability

• Availability heuristic: judging frequency by the ease by which instances come to mind
• Substitute one question for another
• Vivid examples more available
• Largest role to fluency/ease of generating instances
• Retrieval of final instances difficult when need la GE number and causes low ease and thus self rating of worse in that area
• Ease of generating instances is Shteyn 1

13 availability, emotion, and risk

• Affect heuristic: substituting easier question "what do I feel" for harder question "what do I think"
• Availability cascade creates overreaction and craze
• Probability neglect creates gross exaggeration of unlikely events

14: Tom W's speciality

• Using base rate information when don't have any more specific info
• Stereotypes affect rankings of likelihood
• Stereotypes automatic activity of 1
• Predicting by representativeness
• Ignoring base rate or veracity of description
• Even if told that info is false, it is hard to ignore it
• When u know info is false then just stay close to base rates
• Bayes rule for how to combine new info into prior belief

15 Linda less is more

• Incompatibility of heuristics with logic
• Conjunction fallacy: belief that a and b more likely than a
• Confusing plausibility with probability
• Less is more
• Direct comparison makes people more logical

16 causes trump statistics

• Statistical base rate information is often ignored
• But causal base rate information is easy to incorporate such a stereotypes
• Social norm against stereotypes is not cost less
• People who learn about experimental findings ignore surprising conclusions when considering themselves or others close to them
• Surprising statistical facts don't educate
• Surprising anecdotes do stick
• Inferring from specific to general a lot stickier
• Surprising individual cases much more persuasive than statistics

17 regression to the mean

• Rewards from improved performance better than punishments for mistakes
• Attaching causal interpretations to random fluctuations
• Feedback we get is often reversed
• Success is talent plus luck
• Hard for people to understand regression because not causal explanation
• Extreme groups regress to the mean more
• Need control group

18 taming intuitive predictions

• Nonregressive intuition
• We are good at rejecting completely unrelated info but bad at evaluating quality of related evidence
• Need to correct intuition for regression to mean

Part 3 overconfidence

19 illusion of understanding

• Narrative fallacy from taleb
• Halo effect in narratives
• human mind doesn't deal well with nonevents
• Luck plays bigger role in event than in its rebelling
• Easier to construct a story when you know less facts about it
• Don't use word "know" for predicting events
• Future not knowable
• Past easily altered when get surprised; easy to forget what u believed before
• Hindsight bias
• Outcome bias when evaluating decision
• Lucky leaders not punished for taking too much risk

20 the illusion of validity

• Keep feeling like our predictions are accurate even when he got opposite evidence
• Declarations of high confidence just mean that person created coherent story but not that it's true
• Skill can only be demonstrated by consistent performance in multiple trials over time where the role of luck will cancel out
• Illusion of skill in stock picking
• 0 correlation from year to year of same advisors
• Expert not better than non-specialist at forecasting
• Illusion of skill

21 intuitions vs formulas

• Simple algorithm beats experts
• Checklists and simple rules better
• People resist to synthetic algorithms instead of human touch
• Must first collect objective facts about past during interview along concrete scales and then afterwards check intuition
• 6 dimensions which are independent
• Score on each from 1-5
• Collect info on one trait at a time
• Resist desire to hire someone u like more; higher one with best score

22 expert intuition: when can we trust it

• Expertise takes thousands of hours to develop
• Ability to recognize patterns
• Intuitions valid when environment is regular and opportunity to learn over time patterns
• Statistical algos better in noisy environments
• Quicker and better feedback builds skills faster

23 outside view

• Inside view: focusing more on particulars than historical cases and base rate
• Not considering unforeseen circumstances and simply extrapolating from known limited experience too optimistic
• Planning fallacy: unrealistically close to best case scenario
• Must use distributional information from past similar cases as main estimate
• Reference class forecasting
• Irrational exuberance to keep holding on to failing endeavor
• Sunk cost fallacy

24 engine of capitalism

• Optimistic bias
• Entrepreneurial delusions
• Overconfidence
• Premortem as partial solution
• Write speech about disaster of plan before u do it
• Legitimizes doubt and search for threats

Part 4 choices

25 bernoulli's errors

• People don't value uncertain prospects by expected value
• Risk averse decision making
• Pay premium to avoid uncertainty
• Diminishing marginal utility of wealth
• But theory also flawed because happiness is reference dependent and based on recent changes to wealth not just total
• Context matters
• More likely to take risks when all options are bad
• Theory induced blindness

26 prospect theory

• Different preferences if considering gain or loss
• Risk seeking when facing losses
• Need to know reference state and delta
• Weigh losses twice as much as gains

27 endowment effect

• Omission of reference point of current wealth in indifference curve
• Loss aversion creates status quo effect
• Owning a good increases its value
• Huge spread in personal price bid/ask is due to endowment effect
• Price depends on reference point of having or not having the object
• Goods held for exchange different from goods held for use

28 bad events

• Have built in automatic super fast mechanism to recognize threats in system 1/amygdala
• Goals as reference points

29 four fold pattern

• Certainty effect: much more valuable to achieve certainty
• Overweighting small probabilities
• Allay's paradox: we value certainty at a premium and don't just use expected utility
• Prospect theory converts probability percentages into decision weights that rise nonlinearly
• Possibility effect: rare events are over weighted

30 rare events

• Emotion disproportionate to probability
• Denominator neglect due to vivid imagery
• Stating frequency in absolute terms out of 1000 or whatever feels different from probability

31 risk policies

• See gambles as small ones that will repeat to minimize loss aversion
• You win a few, you lose a few
• Broad framing
• Think like a trader
• Better to have risk policy like broad frame instead of deciding each case by case
• Outside view as broad frame for thinking about plans
• Evaluate portfolio only once per quarter

32 keeping score

• Mental accounts irrational
• Disposition effect: more likely to sell winners than losers
• Sunk cost fallacy
• Departure from default is what produces regret
• Favors risk averse choices

33 reversals

• Joint or comparison evaluation places emphasis on different features than standalone single evaluation
• Judgments may be coherent within categories but not across categories
• Joint evaluation Better

34 frames and reality

• Frames are reality
• Moral opinions are based on specific descriptions, not states of reality
• Opt out vs opt in

Part 5 two selves

35 two selves

• Experienced utility
• Duration doesn't affect rating of pain, just intensity
• Peak end average
• Ending of procedure affects rating
• Confusing experience with memory of it
• Memory stores peak and end, not area under curve of pain/time graph
• Duration neglect

36 life as a story

• A story is about significant events and moments, not about time passing so duration neglect makes sense
• Ending is what is remembered
• Desire endings to be good
• Narrative of life
• Caring for people is more for their stories than their feelings
• People choose by memory whether or not to repeat something rather than by actual experience at the time
• You are your remembering self. Your experiencing self is like a stranger to you.

37 experience well being

• Well being experience determined by what we pay attention to

38 thinking about life

• Current mood affects evaluation of life
• Available ideas determine this too
• Nothing in life is as important as it is when ur thinking about it in the moment
• Focusing illusion
• Affective forecasting: assume current state will last forever

conclusions

• Two systems: 1 and 2
• Two species: econs and humans
• Two selves: experiencing and remembering
• Libertarian paternalism through nudges and framing

---