Thinking Fast and Slow

 Thinking Fast and Slow - Daniel Kahneman

Intro

• Biases - systematic errors

• Using heuristics - usually substitute an easier question for a harder one

• System 1 more influential than your experience tells you

Chp 1

• Control of attention shared by 1 and 2

• Limited budget of attention (invisible gorilla)

• System 1 constantly suggesting, feeding info to system 2

• Cannot turn off system 1

• Cognitive illusions

• System 2 cannot substitute for 1, need to compromise

• System 1 - really likes stories, cause-effect, active agents

Chp 2

• 2 is “lazy”

• Measuring cognitive effort by pupil dilation

• Mental effort is distinct from emotional arrousal

• We have a maximum mental effort, above which we “give up”

• High mental effort, hard to pay attention to other things

• Gravitate to least demanding course of action

• Switching from one task to another is effortful

Chp 3

• Self-control and deliberate thought draw on same limited budget of effort

• Flow

• Busy and depleted system 2 - less self control, yield more easily to temptation, more selfish and prejudiced

• All variants of voluntary effort draw at least partly on shared pool of mental energy

• “Ego depletion”

• Motivation

• Overconfidence

• If system 1 involved, conclusion comes first and arguments follow

• Connection between ability to control attention and ability to control emotions

• Stanovich and West (system 1 and 2, or type 1 and 2 process)

• Stanovich says 2 parts to system 2

• Algorithmic and “rational,” or engaged or reflective

Chp 4 - Associative Machine

• Associative activation, cascade

• Coherent, good story, causal story

• Priming effect - relies on unconscious association

• Ideomotor effect - influence of an action by an idea (“florida/old age” experiment), also works in reverse

• Reciprocal priming effects - circular

• Money priming

• “Lady macbeth effect”

Chp 5 - Cognitive Ease

• Fluency level, familiar

• Memory illusion

• Predictable illusions occur if a judgment is based on an impression of cognitive ease or strain. Anything that makes it easier for the associative machine to run smoothly will also bias belief

• Truth illusions (maximize legibility, visual contrast, easy to read, memorable)

• Repetition induces cognitive ease/availability

• Mere exposure effect (usually stronger for stimuli that individual never consciously sees)

• Mednick - creativity is associative memory that works very well, RAT (remote association test)

• Powerful effect of mood on this test

• Happy mood loosens control of system 2

Chp 6 - Norms, surprises, and causes

• System 1 assessing what is normal, maintain model of normality

• “Norm theory” - recruit original episode and interpret in conjunction with it

• Seeing causes and intentions, need for coherence

• Perception of intention and emotions is irresistible (except to some extent for folks on spectrum)

• Experience of freely willed action is different from physical causality

• Paul Blood - inborn readiness to separate physical and intentional causality explains religious beliefs; perceive “world of objects as essentially separate from world of minds”

Chp 7 - machine for jumping to conclusions

• System 1 doesn’t keep track of alternatives it doesn’t select

• System 2 has to decide whether to “unbelieve it”

• Confirmation bias, positive test strategy, search for confirming evidence, is how system 2 tests hypothesis

• Exaggerated emotional coherence (halo effect)

• Suppressed ambiguity

• Halo increased weight of initial impressions

• Decorrelate error!

• What you see is all there is (WYSIATI)

• Asymmetry between ways our mind treat info that is available vs info we don’t have (availability bias)

• System 1 radically insensitive to quality and quantity of information that gives rise to impressions and intuitions

• Consistency and coherence of info more important than completeness;

• Knowing less is often better/easier than knowing more

• Overconfidence

• Framing effects - how present info

• Base rate neglect

Chp 8 - How Judgments Happen

• Directing attention and searching memory

• System 1 able to translate values across dimensions

• Mental shotgun

• Basic assessments

• Friend or foe, approach avoid

• Dominance, attractiveness

• Similarity, representativeness

• System 1 does well with averages but poorly with sums

• Deals with exemplars or prototype

• Almost complete neglect of quantity in certain situations

• Intensity matching

• Prediction by matching not accurate, but comes naturally

• Mental shotgun - system 1 computes more than want or need for particular judgment

Chp 9 - Answering an Easier Question

• Have intuitive feelings and opinions about almost everything

• If question or judgment difficult or not enough time, system 1 find an easier related question (“substitution”)

• Heuristics

• Judgment based on heuristic will be biased in predictable ways

• Present state of mind looms very large when people evaluate their happiness “WYSIATI”

• Affect Heuristic - dominance of conclusions over arguments most pronounced where emotions are involved

• Paul Slovic - let likes and dislikes determine beliefs

• System 2 more of apologist than critic, endorser rather than enforcer

Chp 10 - the Law of Small Numbers

• System 1 inept when faced with “merely statistical” facts

• Sampling effects, misunderstand or ignore the importance of sample size

• Artifacts - observations produced entirely by method of research

• Bias of confidence over doubt

• Focus on the story rather than reliability of results

• Sustaining doubt harder work

• Always looking for causes, make it hard to see randomness (vigilance)

• “To the untrained eye, randomness appears as regularity or tendency to cluster”

• “No such thing as a hot hand” in basketball; satisfies test of randomness

• Illusions of patterns (e.g. gates foundation donating for more ‘small schools’ based on misinterpretation of data [small schools more variability])

Chp 11 - Anchors

• Anchoring effect - occurs when people consider a value for an unknown quantity before estimating that quantity

• People’s judgments influenced by obviously uninformative numbers

• System 2 anchoring - adjustment

• System 1 anchoring - priming

• Anchor and adjust

• Start with anchoring number, then adjust, but usually not far enough, b/c stop at edge of uncertainty

• People adjust less when mental resources depleted

• Priming - evoke info compatible with it

• Powerful anchoring effects in choices about money

• Random anchors can be as effective as informative ones

• Far more suggestible than we would like to be

• Negotiations

• Deliberately think the opposite

• Assume the anchoring effect

Chp 12 - Science of Availability

• Impression of the ease with which instances come to mind

• Salient events, dramatic events, personal experiences

• Fluent retrieval

• If asked for more examples (cognitive strain), tend to believe or estimate lower amounts

• Some resistance to availability bias if more vigilant

• When feel good or powerful, more likely to rely on intuitions, system 1

Chp 13 - availability, emotion, and risk

• Protective actions usually designed based on worst past experience, rather than worst possible

• Tv coverage biased toward novelty and poignancy

• Damasio - emotional evaluations of outcomes

• Haidt - the emotional tail wags the rational dog

• Slovic - people have rich view of risk

• Risk exercise in power

• Availability cascade

• Emotional reaction becomes a story in itself

• Importance judged by fluency and emotional charge

• Hard to deal with small risks: either ignore or drastically overweigh

• Protect public from fear as well as danger

Chp 14 - Tom W’s Specialty

• Base rate

• Stereotype

• Unaffected by size of group (predicting by representativeness, substitution

• Representative heuristic

• Often more accurate than chance guess

• Excessive willingness to predict unlikely events

• Think like statistician vs think like clinician

• Insensitivity to quality of evidence

• Bayesian statistics

Chp 15 - Linda: less is more

• Linda fits idea of “feminist bank teller” than bank teller

• Conjunction fallacy - judge conjunction of events more probable than one of the events

• Substitution of plausability for probability

• Coherence

• System 1 averages rather than adding (sometimes less is more)

• Single evaluation vs joint evaluation

Chp 16 - Causes Trump Statistics

• Statistical base rates

• Causal base rates

• Stereotypes are how we think about categories

• Hard to change mind about human nature

• People tend to exempt themselves from statistics or scientific conclusions (hard to learn from the general)

• But easier to learn from examples, specific to the general

Chp 17 - Regression to the Mean

• Illusory effect of punishment for below standard performance and reward for above standard

• Talent and luck

• Regression strange to the human mind

• It always occurs when correlation between two things less than perfect

• Correlation coefficient

• Correlation and regression two ways of looking at same concept

• Our mind strongly biased to causal explanations

• Often confuse correlation with causation

Chp 18 - taming intuitive predictions

• Insensitive to predictive quality of evidence

• Prediction often matches evaluation

• Correlation b/w two measures equal to proportion of shared factors among their determinants

Chp 19 - The illusion of understanding

• Taleb - narrative fallacy, focus on few striking events

• Halo effect (or negative, horn effect)

• Compelling narrative fosters illusion of inevitability

• Ultimate test of an explanations is whether it would have made the event predictable in advance

• The human mind does not deal well with non-events

• Unlimited ability to ignore our past, believe we understand past, hindsight bias

• Once adopt new view, lose much of your ability to recall what believed before (hindsight bias) - i knew it all along

• Outcome bias (blame and praise agents based on outcome)

• In the presence of randomness, patterns are illusions

Chp 20 - Illusion of validity

• Reluctant to infer the particular from general

• Stock market, picking stocks - an illusion of skill, rewarding luck as if it was skill

• Expert pundits, bad record of predictions

• The more famous the forecaster, the more flamboyant the forecasts

Chp 21 - intuitions vs formulas

• Meehl - clinical vs statistical prediction

• Low validity environments; accuracy of experts matched or exceeded by simple algorithm

• Experts try to be clever, too specific and bold

• Use algorithm except for “broken leg rule”

• Gawande - a checklist manifesto

• Horror/fear of algorithmic mistakes, the cause of a mistake matters

Chp 22 - Expert intuition: when can we trust it

• Klein - Naturalistic Decision Making, how experienced professional develop skills

• Recognition primed decision

• Great facility to learn when to be afraid

• Expertise takes time to develop, need regular environmental feedback

• An environment regular enough to be predictable

• Time to learn these regularities through practice

Chp 23 - the outside view

• Start public discussion by confidentially collecting each person’s judgment

• Proud emphasis on uniqueness of each case

• The planning fallacy - unrealistically close to best case scenarios, could be improved by consulting statistics of similar cases

• Reference class forecasting, overcome base rate neglect

• Optimistic bias, delusional optimism

• Sunk cost fallacy

Chp 24 - the engine of capitalism

• Optimism both a blessing and a risk, largely inherited, less depression, better immune system, feel healthier, promotes action

• Starting small businesses

• People generally feel that they are superior to most others on most desirable traits

• Optimistic risk taking - helps drive economic dynamism

• Competition neglect

• Anchor on our plan, illusion of control

• Cognitive ease vs cognitive strain - judge better than vs worse than

• CFO’s way overconfident in predictions, paid to be knowledgeable, not acceptable to account for all the uncertainty

• Generally a sign of weakness and vulnerability for clinicians to appear unsure

• Premortem, what will we do if this crashes and fails, legitimize doubts

Chp 25 - Bernoulli’s error

• The agent of economic theory is rational, selfish, and his tastes do not change

• Thaler - “Econs”

• Prospect Theory - value the change in status, not just value of outcome, so need to know reference points

• Disbelieving is hard work, and system 2 is often tired

Chp 26 - Prospect Theory

• Dislike losing more than like winning, threats more urgent than opportunities, asymmetrical s-curve

• Diminishing sensitivity

• Loss aversion ratio (1.5 - 2.5), but all bets are off is loss is potentially ruinous

• Doesn’t deal with disappointment, anticipation of regret

Chp 27 - endowment effect

• Indifference map/curve, combination of two goods

• Loss aversion induces a bias that favors status quo

• Owning an object appears to increase its value (endowment effect)

• Held for use vs held for future exchange

• For econs, buying price is irrelevant history, not so for humans

• Decision making under poverty

• Living below one’s reference point

• Always in the “losses”, so improvement is “reduced loss” rather than a gain, all choices between losses

Chp 28 - bad events

• Negativity and escape dominate positivity and approach

• Golfers put more accurately for par than birdie

• Amygdala - threat center; threats can bypass visual cortex (react before “seeing”)

• Some distinctions between good and bad hardwired

• Goals are reference points

• Animals also fight harder to avoid losses than achieve gains

• Loss aversion in law (fairness)

• Asymmetrical effects on well-being

Chp 29 - Fourfold Pattern

• Possibility effect: large impact of changes between 0%-5%

• Overweigh small risks

• Certainty effect: large impact of changes between 95%-100%

• Certainty at a hefty price

• Inadequate sensitivity to intermediate probabilities

• Probability vs decision weight

• Almost completely insensitive to variations of risk among small probabilities

• High probability of gain? Risk averse, fear of disappointment, accept settlement

• Low prob of gain? Hope of large gain, risk seeking, lottery ticket

• Hig prob of loss? Hope to avoid loss, risk seeking, reject favorable settlement

• Low prob of loss? Fear of large loss, risk averse, accept settlement

• Hard to cut losses

Chp 30 - rare events

• Terrorism: availability cascade

• Lotteries and terrorism - same kind of mechanism

• Rare events ignored or overweighted

• Plausability, can you imagine it

• Probability likely to be overestimated if alternative not fully specified

• Vivid and emotional

• Denominator neglect

• Unlikely events more heavily weighted when stated in relative frequencies (1 out of 100) vs abstract risk (1%), people tend to take more “seriously”

• Decisions from global impressions

• Decision from experience - usually don’t overweight, often underweigh

Chp 31 - risk policies

• Gains and losses combined or deconstructed, different preferences

• Narrow (separate decisions) vs broad (single comprehensive decision) framing

• Humans more likely to narrow frame

• Look at gamble as part of bundle of gambles, shield yourself from pain of losses with broad framing

• Creak a risk policy based on broad framing

Chp 32 - keeping score

• Money as proxy for points on a scale

• Mental accounting

• Massive preference for selling winners than losers - disposition effect

• Sunk cost fallacy

• Regret, self-administered punishment

• Stronger reaction to sins of commission vs omission, more regret and blame

• Loss aversion higher for health

• Taboo tradeoff (can’t accept increase in risk)

• Be explicit about anticipation of regret

• Regret and hindsight bias

• Gilbert - people tend to overestimate amount of regret they will feel

Chp 33 - reversals

• Discrepancy between joint and single evaluation

• Joint eval can focus attention on different aspect

• Categories, if eval is across categories, can cause reversal

• Intensity matching

• Hsee’s evaluability hypothesis: somethings not evaluable on their own

• Awards to victims of personal injury were more than twice as large in joint than in single eval

• Joint eval usually broader, but be wary of sales technique or manipulation of joint eval

Chp 34 - frames and reality

• Meaning - associative machinery

• Framing effects - losses hurt more than costs

• Cash discount vs credit surcharge

• Physicians just as susceptible to framing effects

• Frame-bound vs reality-bound

• System 2 has no moral intuitions of its own

• Descriptions vs substance

• Mpg frame is wrong, should be replace by gallons per mile

• Opt-out vs opt-in

• Not how we experience the workings of our mind

Chp 35 - two selves

• Experienced utility (J. Bentham)

• Decision utility (wantability, economics)

• Peak-end rule - avg level of pain at worst moment and end

• Duration neglect - tend to forget/neglect duration

• Experiencing self vs remembering self

• Maximize future memories

• Intensity more impt to memory than duration

Chp 36 - life as a story

• Rules of narratives and plot

• Significant events and memorable moments, progress, gains and losses

• Peak-end rule

• Amnesic vacations

Chp 37 - experience well-being

• Experience sampling (Csikszentmihalyi)

• Day reconstruction method

• Extent of inequality of emotional pain

• Being poor is miserable, being rich man enhance life satisfaction reporting but doesn’t enhance experienced well-being

• Satiation level

Chp 38 - thinking about life

• Gilbert and Wilson: Affective forecasting

• Miswanting - bad choices from affective forecasting

• Focalism - rich source of miswanting

• Mood heuristic one way to answer life-satisfaction question

• Small sample of highly available ideas

• Focusing illusion: substitute small part for the whole (synecdoche)

• Attention to new situations withdrawn over time as it becomes more familiar

• Except for: chronic pain, exposure to loud noise, severe depression

• Bias in favor of goods and experiences that are initially exciting

Conclusions

• Don’t consider humans irrational, have rational capabilities

• Chicago school of economics, faith in human rationality, freedom to choose, milton friedman

• Libertarian paternalism (Nudge, Thaler and Sunstein), defaults, etc

• Marvel of system 1 - maintaining rich and detailed model of world

• System 2 - can’t distinguish b/w info from skills vs heuristic

• Richer language, diagnostic ability