Thinking, Fast and Slow by Daniel Kahneman

Part 1: Two systems

Chapter 1: The characters of the story

• There are two mental agents, System 1 (intuitions, impressions, feelings that require little focus and are automatic) and System 2 (systematic thought that requires deliberate attention). Most of what is eventually thought or done originates in System 1, but System 2 gets the final say
• This book is about recognizing situations in which System 1 errors are likely and avoid them accordingly. It is easier to recognize the mistakes of other rather than our own, due to the illusion of self and the associated self-serving biases

  Chapter 2: Attention and effort

• “Law of least effort” asserts that if there are several ways of achieving the same goal, people will eventually gravitate towards the least demanding course of action
• Task switching and increased temporal/spatial frequency of mental tasks induces strenuous mental effort, engaging System 2, dilating pupils and elevating heart rate

  Chapter 3: The lazy controller

• Self control and cognitive effort are forms of mental work, and studies have shown that when people are simultaneously challenged by a demanding cognitive task and by a temptation, they are more likely to give in to the temptation. System 1 has more influence on behaviour when System 2 is busy
• People who are cognitively busy are more likely to make selfish choices and make superficial judgements in social situations. This too happens to drunk people, or those who are sleep deprived
• The nervous system consumes more glucose than most other parts of the body, and when engaged in a challenging cognitive task, blood glucose levels have been found to drop
• Studies engaged participants in self control tasks that depleted their ego and provided half with lemonade with glucose and the other with splenda. Those who drank the glucose lemonade performed better on subsequent tasks
• Parole judges are more likely to grant parole requests after meals, and more likely to default to rejecting parole a few hours after meals when they are hungry and fatigued
• Oreo test: 4 year old children were given the option to ring a bell to receive 1 oreo, or wait 15 minutes to get 2 oreos. Those who resisted the temptation and waited were found, 10-15 years later, to perform better on cognitive tasks, were less likely to take drugs, and were generally more succesful

  Chapter 4: The associative machine

• Psychological priming: voters voting at polls located in schools are more likely to support pro-education policies. People primed with words associated to money tend to portray more individualistic traits (less likely to help others, less likely to ask for help, more likely to distance themselves from others [literally, people would place chairs further from each other when asked to set up seating arrangements])
• “The lady Macbeth effect”: people who think about actions they are ashamed of are more likely to buy soap, disinfectant, or detergent, some would (dramatically) say to cleanse their body due to their stained soul

  Chapter 5: Cognitive ease

• A happy mood loses control of System 2: when in a good mood, people become more intuitive and creative, but also less vigilant and more prone to logical errors. A good mood is a signal that the environment is safe, so you can let your guard down, whereas a bad mood indicates things are not going well, there may be a threat, and vigilance is required

  Chapter 6: Norns, surprises, and causes

• Norms settle due to repeated exposure to things in our environment, even one exposure settles a norm foundation. Norms are just statistical representations of environmental data - when told “the large mouse climbs over the small elephant’s trunk” we can retrieve the expected values (mean) of the respective animal shapes, as well as their variability in size (i.e the first and second moments of their size distributions)

  Chapter 7: A machine for jumping to conclusions

• When System 2 is engaged elsewhere, or energy is sufficiently depleted as to hinder System 2, we are more susceptible to believe things that are untrue: our System 1 jumps to conclusions. Things like commercials are more effective when watchers are tired and depleted
• The halo effect gives preferential weighting to first impressions. When grading, for example, if you grade one person’s test all at once and they give a good answer on the first question, you’ll be more likely to give them the benefit of the doubt on later mistakes. A better way to address this issue is to do one question at a time, as to decorrelate the grading - this way the grading instances are independent, and errors in grading will average out to zero over enough samples if they are indeed independent

  Chapter 8: How judgments happen

• We evaluate others from the shape of their face, a strong jaw indicates dominance, and a smile indicates friendliness. Political candidates can be meaningfully predicted of success based on their facial features alone
• Those who are politically uninformed and that watch lots of TV tend to be more likely to succumb to System 1 judgements about political candidate based on their facial features
• When given a task, our mind may perform excess computations. For example, when asked to say when vote and note rhyme, participants answered faster than when given the words vote and goat. Though only the audio of the words were provided, people considered the spelling and it slowed down the response - this is known as the mental shotgun

  Chapter 9: Answering an easier question

• A heuristic is a simple procedure that helps find adequate, though often imperfect, answers to difficult questions. It comes from the same root as eureka!
• An experiment was done where students were asked how happy they were these day, and how many dates they had in the last month. The answers in this order were uncorrelated, however when the order of the questions are swapped the answers are significantly correlated. This is because the emotional state with respect to the date question primes the brain, and the answer related to happiness uses a heuristic based on the happiness with respect to the date answer. The present state of mind looms very large when people evaluate their happiness
• Example: Question: how much would you contribute to save an endangered species? Heuristic question: How much emotion do I feel when I think of dying dolphins? Or: Am I happy with my life? -> Am I happy right now?

  Part 2: Heuristics and biases

  Chapter 10: The law of small numbers

• The law of small numbers is when people trust statistical results from small sample sizes, however, small sample sizes result in larger variability and more noise. Transitional researchers in psychology often don’t compute the minimum required sample size to have sufficiently small errors in their studies
• A study by the gates and Melinda foundation found that smaller schools tended to perform better in the top 50, so they funded the creation of smaller schools. What they failed to realize is the variability associated with smaller schools made for some exceptionally strong performers, but small schools also populated the bottom of the list. They fell for the law of small numbers
• We pay more attention to the content of messages than to information about their reliability, and end up with a view of the world that is simpler and more coherent than the data suggest

  Chapter 11: Anchors

• The question: is a redwood more or less than 1200 feet tall? Followed by: what is the average height of a redwood? Results in larger estimates of the average height than if we were to ask if an average redwood is more or less than 180 feet. This is because the user anchors to the value in the first question, and this influences the response of the second
• Real estate agents given house portfolios with either under-valued or over-valued prices were told to come up with a price based on their analyses: the anchoring of the initial price significantly effected their estimates - even though they were confident that they didn’t consider the initial price
• Marketers use the anchoring phenomena. When a sale is labeled, as well as an arbitrary limit to the amount of products you can buy, the average number of purchases of this product will be larger than if the limit was removed (e.g While supplies last!)
• Home sellers take advantage of the anchor effect by setting high prices, it provides the upper hand in negotiations. This effect can be countered, however, by deliberate engagement of System 2 *People are confident that anchors do not affect their judgement, however, anchors prime memory and make information related to it easier to retrieve. System 2 pulls data from this space, unaware it has been affected by the anchoring agent

  Chapter 12: The science of availability

• The availability heuristic is influenced by the following factors:
  • The salience of events adds more weight of them to your memory making them more easily retrievable (e.g divorce among celebrities may seem more common than it actually is since it is reported on so much)
  • Dramatic events add a primacy bias, e.g a recent plane crash will make planes seem unsafe
  • Personal experiences with events will be valued higher on future memory retrieval than had it been, say, reported in the news instead
• Asking someone to name 6 things that they’re competent in or 12 things that they’re competent in leads to different judgements in competence: the more cognitive strain required to retrieve (due to available memory) instances yields a weakened measure in self-competence (if I have to work so hard to get this information, I must not be as competent as I thought)

  Chapter 13: Availability, emotion, and risk

• People with positive views on a subject tend to be able to list more benefits, and those with negative views tend to list more risks. Once they are told the thing that they dislike has more benefits, they perceive it as less risky, if told that the risks are negligible, they develop a more favourable view of the benefits. “The emotional tail wags the rational dog”
• The availability cascade is when a salient event, normally of low risk, catches attention of the public, and thus the attention of the media. This creates a feedback loop between the public and the media, which artificially amplifies the risk of the event. This can result in wasted resources and can be detrimental in some cases (e.g the apple chemical case, where a pesticide was labeled as a low-risk carcinogen and people stopped buying apples and actors testified to congress to remove it - public health was likely impacted negatively as fewer good apples were consumed as a result)
• The combination of probability neglect (giving more weight to numerator (bad events) than the denominator (good events)) with the social mechanisms of availability cascades inevitably leads to exaggeration of minor threats, sometimes with important consequences Chapter 14: Tom W’s speciality
• People tend to neglect base rates when provided specific details about someone. For example, when asking whether it is more likely for someone to be a farmer or librarian, farmers are more common, so we select that as per the base rates. When told whether a timid, organized, quiet loving individual is more likely to be a farmer or librarian, we guess librarian - neglecting the population base rate!
• When System 2 is engaged (induced via frowning), people tend to be more analytical and thus account for the base rates more
• Keys to disciplined Bayesian reasoning:
  • Anchor your judgement of the probability of an outcome on a plausible base rate
  • Question the diagnosticity of your evidence (intuitive impressions tend to exaggerate via the evidence)

Chapter 15: Linda: less is more

• Linda is 31, majored in philosophy, has a passion for social justice and is deeply concerned with discrimination. What is more likely: that Linda is a bank teller, or that Linda is a bank teller who is also active in the feminist movement?
• The second choice is less probable, since feminist bank tellers are a subset of bank tellers, so the probability of feminist bank teller has to be less, yet our intuition tells us otherwise

  Chapter 17: Regression to the mean

• Military flight instructors challenged the idea that rewards are more effective in improving behaviour than punishment. A chief challenged this idea: he argued that after he praised someone, they tended to do worse, and when he yelled at someone for doing something bad they tended to do better after. Due to fluctuations around a mean, unusually bad performances tend to be followed by better performances, and unusually good performances tend to be followed by worse performances. The chief just attributed these results to his punishment, rather to general statistical outcomes
• A business commentator who correctly announces that “the business did better this year because it had done poorly last year” is likely to have short tenure on air, so they attempt to come up with causal stories to explain the results

  Part 3: Overconfidence

  Chapter 19: The illusion of understanding

• Outcome bias: when the outcomes are bad, clients often blame their agents for not seeing the writing on the wall - forgetting that it was written in invisible ink that became legible only after. Actions that seemed prudent in foresight can look irresponsibly negligent in hindsight
• Physicians, financial advisers, and politicians are prone to blame for good decisions that work out badly and are given little credit for successful moves that appear obvious only after the fact
• Stories of success and failure, in the context of business, consistently exaggerate the impact of leadership style and management practices in firm outcomes – hence, their messages are rarely useful
• These stories offer what the human mind needs: a simple message of triumph and failure that identifies clear causes and ignores the determinative power of luck and the inevitability of regression to the mean

  Chapter 20: The illusion of validity

• Confidence is a feeling, which reflects coherence of the information and the cognitive ease of processing it. Declarations of high confidence mainly tell you that an individual has constructed a coherent story in their mind, not necessarily that the story is true
• In the paper “Trading is hazardous to your wealth”, it was shown that on average the most active traders had the poorest results, whereas the investors with the least trading actions earned the highest results
• In the paper “Boys will be boys”, it was shown that men act on their useless ideas significantly more than women (more risk-taking behaviour in males induced by more testosterone), and as a result women achieve better investment results than men
• At an investment advisement firm, the author was provided with investment outcomes over 8 years for the firm’s advisers. What he found was that there was nearly zero correlation on average between year-to-year performance
• The illusion of skill is deeply engrained in the culture of the investment industry. Facts that challenge such basic assumptions threaten to devalue peoples’ livelihoods and self-esteem and are thus not absorbed. When met with statistical proof that their results are due to chance, advisers choose to ignore it as it clashes with their personal impressions from experience

  Chapter 20: Intuitions vs. formulas

  • Statistical models consistently outperform - and are much more consistent - than experts in low validity environments. Example: for medical school admissions, adding interviews at the end and giving interviewers the final say of who’s accepted injects bias and uncertainties into the process, and the quality of candidates generally suffers as a result • Questions in an interview should be as objective as possible. It has been shown that scores accumulated from these more objective results can make useful predictions, whereas results based purely on somewhat unstructured interviews aren’t predictive of future success. Objective questions are also better guides for intuitive judgements - which perform on par with the scoring mechanism

  Chapter 22: Expert intuition: when can we trust it?

• Intuition cannot be trusted in the absence of stable regularities in the environment. A chess player’s intuition is valid, due to the predictable environment, whereas a stock picker’s intuition is likely false, due to the high dimensionality and unpredictability of the market
• Experts who receive short term feedback are also more likely to develop intuitions, however with long term forecasts the feedback comes further away from the actions, and thus mapping that back to decision making is more difficult

  Chapter 23: The outside view

• Planning fallacy: when plans and forecasts are unrealistically close to best-case scenarios, and could be improved by consulting the statistics of similar cases
• To avoid the planning fallacy in a new project:
  • First find your reference class to compare against
  • Obtain the statistics of outcomes from the reference class. Use this as a base rate for success
  • Slightly adjust the base rate based on your projects circumstances

Chapter 24: The engine of capitalism

• From the finance industry to the medical industry, it is a sign of weakness and vulnerability for experts to appear unsure. Confidence is valued over uncertainty.
• Clinicians who were completely certain of their diagnoses are wrong 40% of the time. Confidence intervals from CFOs on stock performance are more than 3 times likely for outcomes to lie outside their 80% confidence intervals (i.e surprises)

  Part 4: Choices

  Chapter 26: Prospect theory

• Losses hurt more than gains, and gambles we are willing to take depend on our reference point. Losing 100$ when you have 100$ hurts a lot more than losing 100$ when you have 1000$.

  Chapter 27: The endowment effect

• People who buy products “for use”, rather than “for exchange”, tend to value losing the product as higher than if they were to buy it, e.g willing to pay $500 for a sold out concert ticket, but willing to sell it only for much more. This is an example of loss aversion, as the product holds sentimental value as well

  Chapter 28: Bad events

• Emotionally loaded words quickly attract attention, and bad words (war, crime) attract attention faster than happy words (peace, love)
• Professional golfers rate of success for putts for par and birdies differs by nearly 4%: golfers perform slightly better on par putts to avoid the loss ensued by a bogey

  Chapter 29: The fourfold pattern

• The decision weights that people assign to outcomes are not identical to the probabilities of those outcomes: improbable outcomes are overweighted (the possibility effect), and near certain outcomes are underweighted relative to absolute certainty (the certainty effect)
• A 0 to 5% chance of winning a million dollars is valued higher than a 5 to 10% chance (possibility effect). Likewise, a 95-100% chance is valued higher than a 65-70% chance of winning - certainty is valued more

  Chapter 30: Rare events

• Denominator neglect: for low probabilities, people tend to over emphasize the numerator and neglect the denominator. A vaccine that has a 0.001% chance to disable someone feels different than 1 in 100,000 children will be permanently disabled, because now we anchor to the 1 child and overemphasize the likelihood due to the emotional provocation
• Researchers have leveraged this fact, one study stated “1000 homicides a year are committed by seriously mentally ill individuals”, which corresponds to a 0.00036% likelihood (1000/273mill). The researchers use the former to elicit a fear response to receive more funding for mental health research and services

  Chapter 32: Keeping score

• There exists a preference for selling winners rather than losers in the stock market - this is a bias towards feeling pleasure and the avoidance of feeling the pain of a loss. Yet, studies show it is slightly better to hold on to winners and let go of losers, selling winners is pleasurable but there is a cost to that pleasure
• The sunk-cost fallacy is investing additional resources into a losing account because you have already committed to the account, when better investments are available elsewhere
• In the context of gambling: people expect to be happier if they gamble and win, than if they refrained from gambling and got the same amount
• People generally anticipate more regret than they will actually experience. You should not put too much weight on regret; even if you have some, it will hurt less than you now think

  Chapter 34: Frames and reality

• Individuals were provided a scenario in which they were given $50, and told they could either keep 20$, or lose 30$ - equivalent statements. There was a tendency to choose to keep 20$, as loss aversion makes the keeping frame more attractive
• There was more amygdala (associated with emotional arousal) activation in subjects who produced an immediate tendency to approach the sure gain (keep 20) or avoid the sure loss (lose 30)
• Those who resisted the keep 20, had more activity in the anterior cingulate, which is associated with conflict and self-control
• Those who showed no preference in outcome, thus being more rational, showed enhanced activity in a frontal area of the brain that is implicated in combining emotion and reasoning to guide decisions
• Medical professionals were more likely to suggest a surgery with a 90% survival rate than a surgery with a 10% mortality rate, even though these are equivalent statements

  Part 5: Two selves

  Chapter 35: Two selves

• When rating the subjective pain experienced during an experience, rating done in the present don’t equate to the retrospective rating. Experiences with a prolonged painful duration, yet that gradually decrease in pain over time, are reported to be less painful than a quick experience that ends poorly
• We can have a vastly net positive experience, but, if it ends on a poor note we’re quick to say that the experience was ruined, discounting all the positive experiences prior. This is confusing our experiences with the memory of them, and weighting the experience in our memory higher than the actual experience itself
• Our memories do not remember based on a collective sum of our experience (i.e where duration of an experience would increase our perceived pain/pleasure), but rather a skewed average of the experience, where the most recent peak of pleasure and pain are over-valued.
• Our memory neglects the duration of episodes, and thus does not serve our preferences for long pleasure and short pains

  Chapter 36: Life as a story

• When on vacation, our remembering self is more engaged, for example by taking pictures to engage with in the future (normally putting too much weight on the future self at the expense of the present self)
• When asked how much you would pay for your last vacation if you hadn’t remembered it (i.e you got to experience it in the present but remembering self forgets), people would likely value the vacation less
• Now imagine you undergo painful surgery in which you will be in pain and conscious, but you can take an amnesia pill to forget the experience - how does that sound? The preference goes to the remembering self. It’s odd, the present experience self is undervalued and somewhat a stranger, and we are our remembering selves

  Chapter 37: Experienced well-being

• To get pleasure from eating, you need to notice that you are doing it. It was found that women in France perceive eating as more pleasurable due to eating being twice as focal, despite eating for the same amount of time, than American women do (likely since American women spend time doing other things while eating which dilutes the pleasure)

  Chapter 38: Thinking about life

• The focusing illusion: nothing in life is as important as you think it is when you are thinking about it. Any aspect of life to which attention is directed will loom large in a global evaluation
• Miswanting describes bad choices that arise from errors of affective forecasting. It makes us prone to exaggerate the effect of significant purchases or changed circumstances on our future well-being (e.g moving to this warmer place will make me happier)