Book Summary: Thinking, Fast and Slow, by

Learn the key points in minutes.

Book Rating by Shortform Readers: 4.8 (175 reviews)

This is a preview of the Shortform book summary of Thinking, Fast and Slow by Daniel Kahneman. Read the full comprehensive summary at Shortform.

1-Page Book Summary of Thinking, Fast and Slow

Thinking, Fast and Slow concerns a few major questions: how do we make decisions? And in what ways do we make decisions poorly?

The book covers three areas of Daniel Kahneman’s research: cognitive biases, prospect theory, and happiness.

System 1 and 2

Kahneman defines two systems of the mind.

System 1: operates automatically and quickly, with little or no effort, and no sense of voluntary control

  • Examples: Detect that one object is farther than another; detect sadness in a voice; read words on billboards; understand simple sentences; drive a car on an empty road.

System 2: allocates attention to the effortful mental activities that demand it, including complex computations. Often associated with the subjective experience of agency, choice and concentration

  • Examples: Focus attention on a particular person in a crowd; exercise faster than is normal for you; monitor your behavior in a social situation; park in a narrow space; multiple 17 x 24.

System 1 automatically generates suggestions, feelings, and intuitions for System 2. If endorsed by System 2, intuitions turn into beliefs, and impulses turn into voluntary actions.

System 1 can be completely involuntary. You can’t stop your brain from completing 2 + 2 = ?, or from considering a cheesecake as delicious. You can’t unsee optical illusions, even if you rationally know what’s going on.

A lazy System 2 accepts what the faulty System 1 gives it, without questioning. This leads to cognitive biases. Even worse, cognitive strain taxes System 2, making it more willing to accept System 1. Therefore, we’re more vulnerable to cognitive biases when we’re stressed.

Because System 1 operates automatically and can’t be turned off, biases are difficult to prevent. Yet it’s also not wise (or energetically possible) to constantly question System 1, and System 2 is too slow to substitute in routine decisions. We should aim for a compromise: recognize situations when we’re vulnerable to mistakes, and avoid large mistakes when the stakes are high.

Cognitive Biases and Heuristics

Despite all the complexities of life, notice that you’re rarely stumped. You rarely face situations as mentally taxing as having to solve 9382 x 7491 in your head.

Isn’t it profound how we can make decisions without realizing it? You like or dislike people before you know much about them; you feel a company will succeed or fail without really analyzing it.

When faced with a difficult question, System 1 substitutes an easier question, or the heuristic question. The answer is often adequate, though imperfect.

Consider the following examples of heuristics:

  • Target question: Is this company’s stock worth buying? Will the price increase or decrease?
    • Heuristic question: How much do I like this company?
  • Target question: How happy are you with your life?
    • Heuristic question: What’s my current mood?
  • Target question: How far will this political candidate get in her party?
    • Heuristic question: Does this person look like a political winner?

These are related, but imperfect questions. When System 1 produces an imperfect answer, System 2 has the opportunity to reject this answer, but a lazy System 2 often endorses the heuristic without much scrutiny.

Important Biases and Heuristics

Confirmation bias: We tend to find and interpret information in a way that confirms our prior beliefs. We selectively pay attention to data that fit our prior beliefs and discard data that don’t.

“What you see is all there is”: We don’t consider the global set of alternatives or data. We don’t realize the data that are missing. Related:

  • Planning fallacy: we habitually underestimate the amount of time a project will take. This is because we ignore the many ways things could go wrong and visualize an ideal world where nothing goes wrong.
  • Sunk cost fallacy: we separate life into separate accounts, instead of considering the global account. For example, if you narrowly focus on a single failed project, you feel reluctant to cut your losses, but a broader view would show that you should cut your losses and put your resources elsewhere.

Ignoring reversion to the mean: If randomness is a major factor in outcomes, high performers today will suffer and low performers will improve, for no meaningful reason. Yet pundits will create superficial causal relationships to explain these random fluctuations in success and failure, observing that high performers buckled under the spotlight, or that low performers lit a fire of motivation.

Anchoring: When shown an initial piece of information, you bias toward that information, even if it’s irrelevant to the decision at hand. For instance, in one study, when a nonprofit requested $400, the average donation was $143; when it requested $5, the average donation was $20. The first piece of information (in this case, the suggested donation) influences our decision (in this case, how much to donate), even though the suggested amount shouldn’t be relevant to deciding how much to give.

Representativeness: You tend to use your stereotypes to make decisions, even when they contradict common sense statistics. For example, if you’re told about someone who is meek and keeps to himself, you’d guess the person is more likely to be a librarian than a construction worker, even though there are far more of the latter than the former in the country.

Availability bias: Vivid images and stronger emotions make items easier to recall and are overweighted. Meanwhile, important issues that do not evoke strong emotions and are not easily recalled are diminished in importance.

Narrative fallacy: We seek to explain events with coherent stories, even though the event may have occurred due to randomness. Because the stories sound plausible to us, it gives us unjustified confidence about predicting the future.

Prospect Theory

Traditional Expected Utility Theory

Traditional “expected utility theory” asserts that people are rational agents that calculate the utility of each situation and make the optimum choice each time.

If you preferred apples to bananas, would you rather have a 10% chance of winning an apple, or 10% chance of winning a banana? Clearly you’d prefer the former.

The expected utility theory explained cases like these, but failed to explain the phenomenon of risk aversion, where in some situations a lower-expected-value choice was preferred.

Consider: Would you rather have an 80% chance of gaining $100 and a 20% chance to win $10, or a certain gain of $80?

The expected value of the former is greater (at $82) but most people choose the latter. This makes no sense in classic utility theory—you should be willing to take a positive expected value gamble every time.

Furthermore, it ignores how differently we feel in the case of gains and losses. Say Anthony has $1 million and Beth has $4 million. Anthony gains $1 million and Beth loses $2 million, so they each now have $2 million. Are Anthony and Beth equally happy?

Obviously not - Beth lost, while Anthony gained. Puzzling with this concept led Kahneman to develop prospect theory.

Prospect Theory

The key insight from the above example is that evaluations of utility are not purely dependent on the current state. Utility depends on changes from one’s reference point. Utility is attached to changes of wealth, not states of wealth. And losses hurt more than gains.

Prospect theory can be summarized in 3 points:

1. When you evaluate a situation, you compare it to a neutral reference point.

  • Usually this refers to the status quo you currently experience. But it can also refer to an outcome you expect or feel entitled to, like an annual raise. When you don’t get something you expect, you feel crushed, even though your status quo hasn’t changed.

2. Diminishing marginal utility applies to changes in wealth (and to sensory inputs).

  • Going from $100 to $200 feels much better than going from $900 to $1,000. The more you have, the less significant the change feels.

3. Losses of a certain amount trigger stronger emotions than a gain of the same amount.

  • Evolutionarily, the organisms that treated threats more urgently than opportunities tended to survive and reproduce better. We have evolved to react extremely quickly to bad news.

There are a few practical implications of prospect theory.

Possibility Effect

Consider which is more meaningful to you:

  • Going from a 0% chance of winning $1 million to 5% chance
  • Going from a 5% chance of winning $1 million to 10% chance

Most likely you felt better about the first than the second. The mere possibility of winning something (that may still be highly unlikely) is overweighted in its importance. We fantasize about small chances of big gains. We obsess about tiny chances of very bad outcomes.

Certainty Effect

Now consider how you feel about these options on the opposite end of probability:

  • In a surgical procedure, going from a 90% success rate to 95% success rate.
  • In a surgical procedure, going from a 95% success rate to 100% success rate

Most likely, you felt better about the second than the first. Outcomes that are almost certain are given less weight than their probability justifies. 95% success rate is actually fantastic! But it doesn’t feel this way, because it’s not 100%.

Status Quo Bias

You like what you have and don’t want to lose it, even if your past self would have been indifferent about having it. For example, if your boss announces a raise, then ten minutes later said she made a mistake and takes it back, this is experienced as a dramatic loss. However, if you heard about this happening to someone else, you likely would see the change as negligible.

Framing Effects

The context in which a decision is made makes a big difference in the emotions that are invoked and the ultimate decision. Even though a gain can be logically equivalently defined as a loss, because losses are so much more painful, different framings may feel very different.

For example, a medical procedure with a 90% chance of survival sounds more appealing than one with a 10% chance of mortality, even though they’re identical.

Happiness and the Two Selves

The new focus of Kahneman’s recent research is happiness. Happiness is a tricky concept. There is in-the-moment happiness, and there is overall well being. There is happiness we experience, and happiness we remember.

Kahneman presents two selves:

  • The experiencing self: the person who feels pleasure and pain, moment to moment. This experienced utility would best be assessed by measuring happiness over time, then summing the total happiness felt over time. (In calculus terms, this is integrating the area under the curve.)
  • The remembering self: the person who reflects on past experiences and evaluates it overall.

The remembering self factors heavily in our thinking. **After a moment has passed,...

Want to learn the rest of Thinking, Fast and Slow in 21 minutes?

Unlock the full book summary of Thinking, Fast and Slow by signing up for Shortform.

Shortform summaries help you learn 10x faster by:

  • Being 100% comprehensive: you learn the most important points in the book
  • Cutting out the fluff: you don't spend your time wondering what the author's point is.
  • Interactive exercises: apply the book's ideas to your own life with our educators' guidance.

Here's a preview of the rest of Shortform's Thinking, Fast and Slow summary:

Thinking, Fast and Slow Summary Part 1-1: Two Systems of Thinking

We believe we’re being rational most of the time, but really much of our thinking is automatic, done subconsciously by instinct. Most impressions arise without your knowing how they got there. Can you pinpoint exactly how you knew a man was angry from his facial expression, or how you could tell that one object was farther away than another, or why you laughed at a funny joke?

This becomes more practically important for the decisions we make. Often, we’ve decided what we’re going to do before we even realize it. Only after this subconscious decision does our rational mind try to justify it.

The brain does this to save on effort, substituting easier questions for harder questions. Instead of thinking, “should I invest in Tesla stock? Is it priced correctly?” you might instead think, “do I like Tesla cars?” The insidious part is, you often don’t notice the substitution. This type of substitution produces systematic errors, also called biases. We are blind to our blindness.

System 1 and System 2 Thinking

In Thinking, Fast and Slow, Kahneman defines two systems of the mind:

System 1: operates automatically and quickly, with little or no effort, and no sense of voluntary control

  • Examples: Detect that one object is farther than another; detect sadness in a voice; read words on billboards; understand simple sentences; drive a car on an empty road.

System 2: allocates attention to the effortful mental activities that demand it, including complex computations. Often associated with the subjective experience of agency, choice and concentration

  • Examples: Focus attention on a particular person in a crowd; exercise faster than is normal for you; monitor your behavior in a social situation; park in a narrow space; multiple 17 x 24.

Properties of System 1 and 2 Thinking

System 1 can be completely involuntary. You can’t stop your brain from completing 2 + 2 = ?, or from considering a cheesecake as delicious. You can’t unsee optical illusions, even if you...

Thinking, Fast and Slow Summary Part 1-2: System 2 Has a Maximum Capacity

System 2 thinking has a limited budget of attention - you can only do so many cognitively difficult things at once.

This limitation is true when doing two tasks at the same time - if you’re navigating traffic on a busy highway, it becomes far harder to solve a multiplication problem.

This limitation is also true when one task comes after another - depleting System 2 resources earlier in the day can lower inhibitions later. For example, a hard day at work will make you more susceptible to impulsive buying from late-night infomercials. This is also known as “ego depletion,” or the idea that you have a limited pool of willpower or mental resources that can be depleted each day.

All forms of voluntary effort - cognitive, emotional, physical - seem to draw at least partly on a shared pool of mental energy.

  • Stifling emotions during a sad film worsens physical stamina later.
  • Memorizing a list of seven digits makes subjects more likely to yield to more decadent desserts.

Differences in Demanding Tasks

The law of least effort states that “if there are several ways of achieving the same goal, people will eventually gravitate to the least demanding course of action.”

What makes some cognitive operations more demanding than others? Here are a few examples:

  • Holding in memory several ideas that require separate actions (like memorizing a supermarket shopping list). System 1 is not capable of dealing with multiple distinct topics at once.
  • Obeying an instruction that overrides habitual responses, thus requiring “executive control.”
    • This covers cognitive, emotional, and physical impulses of all kinds.
  • Switching between tasks.
  • Time pressure.

In the lab, the strain of a cognitive task can be measured by pupil size - the harder the task, the more the pupil dilates, in real time. Heart rate also increases.

Kahneman cites one particular task as the limit of what most people can do in the lab, dilating the pupil by 50% and increasing heart rate by 7bpm. The task is “Add-3”:

  • Write several 4 digit numbers on separate...

Thinking, Fast and Slow Summary Part 1-3: System 1 is Associative

Think of your brain as a vast network of ideas connected to each other. These ideas can be concrete or abstract. The ideas can involve memories, emotions, and physical sensations.

When one node in the network is activated, say by seeing a word or image, it automatically activates its surrounding nodes, rippling outward like a pebble thrown in water.

As an example, consider the following two words:

“Bananas Vomit”

Suddenly, within a second, reading those two words may have triggered a host of different ideas. You might have pictured yellow fruits; felt a physiological aversion in the pit of your stomach; remembered the last time you vomited; thought about other diseases - all done automatically without your conscious control.

The evocations can be self-reinforcing - a word evokes memories, which evoke emotions, which evoke facial expressions, which evoke other reactions, and which reinforce other ideas.

Links between ideas consist of several forms:

  • Cause → Effect
  • Belonging to the Same Category (lemon → fruit)
  • Things to their properties (lemon → yellow, sour)

Association is Fast and Subconscious

In the next exercise, you’ll be shown three words. Think of a new word that fits with each of the three words in a phrase.

Here are the three words:

cottage Swiss cake

Ready?

A common answer is “cheese.” Cottage cheese, Swiss cheese, and cheesecake. You might have thought of this quickly, without really needing to engage your brain deeply.

The next exercise is a little different. You’ll be given two sets of three words. Within seconds, decide which one feels better, without defining the new word:

sleep mail switch
salt deep foam

Ready?

You might have found that the second one felt better. Isn’t that odd? There is a very faint signal from the...

What Our Readers Say

This is the best summary of Thinking, Fast and Slow I've ever read. I learned all the main points in just 20 minutes.

Learn more about our summaries →

Thinking, Fast and Slow Summary Part 1-4: How We Make Judgments

System 1 continuously monitors what’s going on outside and inside the mind and generates assessments with little effort and without intention. The basic assessments include language, facial recognition, social hierarchy, similarity, causality, associations, and exemplars.

  • In this way, you can look at a male face and consider him competent (for instance, if he has a strong chin and a slight confident smile).
  • The survival purpose is to monitor surroundings for threats.

However, not every attribute of the situation is measured. System 1 is much better at determining comparisons between things and the average of things, not the sum of things. Here’s an example:

In the below picture, try to quickly determine what the average length of the lines is.

alt_text

Now try to determine the sum of the length of the lines. This is less intuitive and requires System 2.

Unlike System 2 thinking, these basic assessments of System 1 are not impaired when the observer is cognitively busy.

In addition to basic assessments: System 1 also has two other characteristics:

1) Translating Values Across Dimensions, or Intensity Matching

System 1 is good at comparing values on two entirely different scales. Here’s an example.

Consider a minor league baseball player. Compared to the rest of the population, how athletic is this player?

Now compare your judgment to a different scale: If you had to convert how athletic the player is into a year-round weather temperature, what temperature would you choose?

Just as a minor league player is above average but not the top tier, the temperature you chose might be something like 80 Fahrenheit.

As another example, consider comparing crimes and punishments, each expressed as musical volume. If a soft-sounding crime is followed by a piercingly loud punishment, then this means a large mismatch that might indicate injustice.

2) Mental Shotgun

System 1 often carries out more computations than are needed. Kahneman...

Thinking, Fast and Slow Summary Part 1-5: Biases of System 1

Putting it all together, we are most vulnerable to biases when:

  • System 1 forms a narrative that conveniently connects the dots and doesn’t express surprise.
  • Because of the cognitive ease by System 1, System 2 is not invoked to question the data. It merely accepts the conclusions of System 1.

In day-to-day life, this is acceptable if the conclusions are likely to be correct, the costs of a mistake are acceptable, and if the jump saves time and effort. You don’t question whether to brush your teeth each day, for example.

In contrast, this shortcut in thinking is risky when the stakes are high and there’s no time to collect more information, like when serving on a jury, deciding which job applicant to hire, or how to behave in an weather emergency.

We’ll end part 1 with a collection of biases.

What You See is All There Is: WYSIATI

When presented with evidence, especially those that confirm your mental model, you do not question what evidence might be missing. System 1 seeks to build the most coherent story it can - it does not stop to examine the quality and the quantity of information.

In an experiment, three groups were given background to a legal case. Then one group was given just the plaintiff’s argument, another the defendant’s argument, and the last both arguments.

Those given only one side gave a more skewed judgment, and were more confident of their judgments than those given both sides, even though they were fully aware of the setup.

We often fail to account for critical evidence that is missing.

Halo Effect

If you think positively about something, it extends to everything else you can think about that thing.

Say you find someone visually attractive and you like this person for that reason. As a result, you are more likely to find her intelligent or capable, even if you have no evidence of this. Even further, you tend to like intelligent people, and now that you think she’s intelligent, you like her better than you did before, causing a feedback loop.

In other words, your emotional response fills in...

Thinking, Fast and Slow Summary Part 2: Heuristics and Biases | 1: Statistical Mistakes

Kahneman transitions to Part 2 from Part 1 by explaining more heuristics and biases we’re subject to.

The general theme of these biases: we prefer certainty over doubt. We prefer coherent stories of the world, clear causes and effects. Sustaining incompatible viewpoints at once is harder work than sliding into certainty. A message, if it is not immediately rejected as a lie, will affect our thinking, regardless of how unreliable the message is.

Furthermore, we pay more attention to the content of the story than to the reliability of the data. We prefer simpler and coherent views of the world and overlook why those views are not deserved. We overestimate causal explanations and ignore base statistical rates. Often, these intuitive predictions are too extreme, and you will put too much faith in them.

This chapter will focus on statistical mistakes - when our biases make us misinterpret statistical truths.

The Law of Small Numbers

The smaller your sample size, the more likely you are to have extreme results. When you have small sample sizes, do NOT be misled by outliers.

A facetious example: in a series of 2 coin tosses, you are likely to get 100% heads. This doesn’t mean the coin is rigged.

In this case, the statistical mistake is clear. But in more complicated scenarios, outliers can be deceptive.

Case 1: Cancer Rates in Rural Areas

A study found that certain rural counties in the South had the lowest rates of kidney cancer. What was special about these counties - something about the rigorous hard work of farming, or the free open air?

The same study then looked at the counties with the highest rates of kidney cancer. Guess what? They were also rural areas!

We can infer that the fresh air and additive-free food of a rural lifestyle explain low rates of kidney cancer; we can also infer that the poverty and high-fat diet of a rural lifestyle explain high rates of kidney cancer. But we can’t have it both ways. It doesn’t make sense to attribute both low and high cancer rates to a rural lifestyle.

If it’s not...

Thinking, Fast and Slow Summary Part 2-2: Anchors

Anchoring describes the bias where you depend too heavily on an initial piece of information when making decisions.

In quantitative terms, when you are exposed to a number, then asked to estimate an unknown quantity, the initial number affects your estimate of the unknown quantity. Surprisingly, this happens even when the number has no meaningful relevance to the quantity to be estimated.

Examples of anchoring:

  • Students are split into two groups. One group is asked if Gandhi died before or after age 144. The other group is asked if Gandhi died before or after age 32. Both groups are then asked to estimate what age Gandhi actually died at. The first group, who were asked about age 144, estimated a higher age of death than students who were asked about age 32, with a difference in average guesses of over 15 years.
  • Students were shown a wheel of fortune game that had numbers on it. The game was rigged to show only the numbers 10 or 65. The students were then asked to estimate the % of African nations in the UN. The average estimates came to 25% and 45%, based on whether they were shown 10 or 65, respectively.
  • A nonprofit requested different amounts of donations in its requests. When it requested $400, the average donation was $143; when requesting $5, the average donation was $20.
  • In online auctions, the Buy Now prices serve as anchors for the final price.
  • Arbitrary rationing, like supermarkets with “limit 12 per person,” makes people buy more cans, compared to when there’s no limit. (Shortform note: this might also be confounded as a signal of demand, indicating quality or scarcity.)

Note how in several examples above, the number given is not all that relevant to the question at hand. The wheel of fortune number has nothing to do with African countries in the UN; the requested donation size should have little effect on how much you personally want to donate. But it still has an effect.

Sometimes, the anchor works because you infer the number is given for a reason, and it’s a reasonable place to adjust from. But again...

Thinking, Fast and Slow Summary Part 2-3: Availability Bias

When trying to answer the question “what do I think about X?,” you actually tend to think about the easier but misleading questions, “what do I remember about X, and how easily do I remember it?” The more easily you remember something, the more significant you perceive what you’re remembering to be. In contrast, things that are hard to remember are lowered in significance.

More quantitatively, when trying to estimate the size of a category or the frequency of an event, you instead use the heuristic: how easily do the instances come to mind? Whatever comes to your mind more easily is weighted as more important or true. This is the availability bias.

This means a few things:

  • Items that are easier to recall take on greater weight than they should.
  • When estimating the size of a category, like “dangerous animals,” if it’s easy to retrieve items for a category, you’ll judge the category to be large.
  • When estimating the frequency of an event, if it’s easy to think of examples, you’ll perceive the event to be more frequent.

In practice, this manifests in a number of ways:

  • Events that trigger stronger emotions (like terrorist attacks) are more readily available than events that don’t (like diabetes), causing you to overestimate the importance of the more provocative events.
  • More recent events are more available than past events, and are therefore judged to be more important.
  • More vivid, visual examples are more available than mere words. For instance, it’s easier to remember the details of a painting than it is to remember the details of a passage of text. Consequently, we often value visual information over verbal.
  • Personal experiences are more available than statistics or data.
    • Famously, spouses were asked for their % contribution to household tasks. When you add both spouses’ answers, the total tends to be more than 100% - for instance, each spouse believes they contribute 70% of household tasks. Because of availability bias, one spouse primarily sees the work they had done and not their spouse’s...

Why are Shortform Summaries the Best?

We're the most efficient way to learn the most useful ideas from a book.

Cuts Out the Fluff

Ever feel a book rambles on, giving anecdotes that aren't useful? Often get frustrated by an author who doesn't get to the point?

We cut out the fluff, keeping only the most useful examples and ideas. We also re-organize books for clarity, putting the most important principles first, so you can learn faster.

Always Comprehensive

Other summaries give you just a highlight of some of the ideas in a book. We find these too vague to be satisfying.

At Shortform, we want to cover every point worth knowing in the book. Learn nuances, key examples, and critical details on how to apply the ideas.

3 Different Levels of Detail

You want different levels of detail at different times. That's why every book is summarized in three lengths:

1) Paragraph to get the gist
2) 1-page summary, to get the main takeaways
3) Full comprehensive summary and analysis, containing every useful point and example

Thinking, Fast and Slow Summary Part 2-4: Representativeness

Read the following description of a person.

Tom W. is meek and keeps to himself. He likes soft music and wears glasses. Which profession is Tom W. more likely to be? 1) Librarian. 2) Construction worker.

If you picked librarian without thinking too hard, you used the representativeness heuristic - you matched the description to the stereotype, while ignoring the base rates.

Ideally, you should have examined the base rate of both professions in the male population, then adjusted based on his description. Construction workers outnumber librarians by 10:1 in the US - there are likely more shy construction workers than all librarians!

More generally, the representativeness heuristic describes when we estimate the likelihood of an event by comparing it to an existing prototype in our minds - matching like to like. But just because something is plausible does not make it more probable.

The representativeness heuristic is strong in our minds and hard to overcome. In experiments, even when people receive data about base rates (like about the proportion of construction workers to librarians), people tend to ignore this information, trusting their stereotype matching more than actual statistics.

(Shortform note: even after reading this, you might think - but what about self-selection? Don’t meek people tend to seek library jobs and stay away from construction jobs? Isn’t it possible that all the shy librarians outnumber all the shy construction workers, even though there are 10 times more construction workers than librarians? This just goes to show how entrenched the representativeness heuristic is—you seek to justify your stereotype rather than looking at the raw data.)

Here’s another example:

Someone on the subway is reading the New York Times. Is the stranger more likely to have a PhD, or to not have a college degree?

Again, by pure number of people, there are far more people in the latter group than the former.

Why Do We Suffer from the Representativeness Heuristic?

Representativeness is used because System 1 desires coherence,...

Thinking, Fast and Slow Summary Part 2-5: Overcoming the Heuristics

As we’ve been discussing, the general solution to overcoming statistical heuristics is by estimating the base probability, then making adjustments based on new data. Let’s work through an example.

Julie is currently a senior in a state university. She read fluently when she was four years old. What is her grade point average (GPA)?

People often compute this using intensity matching and representativeness, like so:

  • Reading fluently at 4 puts her at, say, the 90th percentile of all kids.
  • The 90th percentile GPA is somewhere around a 3.9.
  • Thus Julie likely has a 3.9 GPA.

Notice how misguided this line of thinking is! People are predicting someone’s academic performance 2 decades later based on how they behaved at 4. System 1 pieces together a coherent story about a smart kid becoming a smart adult.

The proper way to answer questions like these is as follows:

  • Start by estimating the average GPA - this is the base data if you had no information about the student whatsoever. Say this is 3.0.
  • Determine the GPA that matches your impression of the evidence. In this case, you might think someone who was fluent when she was 4 years old would have a GPA of 3.9.
  • Estimate the correlation between your evidence and GPA. Do you think fluent reading at 4 years old is 100% correlated with academic success in college? Likely not—there are factors that affect reading age that aren’t relevant to college GPA, and vice versa. Is it 10% correlated? 50%?
  • FInally, depending on the strength of your correlation, move from the average GPA from the first step to your estimated GPA in the second step. If you believe the correlation is just 10%, then you would move 10% from 3.0 to 3.9, ending up at around 3.1.

This methodical approach is generalizable to any similar prediction task. It avoids overly extreme results from intuition, instead using base rates and assessing the quality of information. It allows for regression toward the mean (eg replace “average GPA and student’s GPA” with “day 1 golf score and day 2 golf score”).

Are Smart...

Thinking, Fast and Slow Summary Part 3: Overconfidence | 1: Flaws In Our Understanding

Part 3 explores biases that lead to overconfidence. With all the heuristics and biases described above working against us, when we construct satisfying stories about the world, we vastly overestimate how much we understand about the past, present, and future.

The general principle of the biases has been this: we desire a coherent story of the world. This comforts us in a world that may be largely random. If it’s a good story, you believe it.

Insidiously, the fewer data points you receive, the more coherent the story you can form. You often don’t notice how little information you actually have and don’t wonder about what is missing. You focus on the data you have, and you don’t imagine all the events that failed to happen (the nonevents). You ignore your ignorance.

And even if you’re aware of the biases, you are nowhere near immune to them. Even if you’re told that these biases exist, you often exempt yourself for being smart enough to avoid them.

The ultimate test of an explanation is whether it can predict future events accurately. This is the guideline by which you should assess the merits of your beliefs.

Narrative Fallacy

We desire packaging up a messy world into a clean-cut story. It is unsatisfying to believe that outcomes are based largely on chance, partially because this makes the future unpredictable. But in a world of randomness, regular patterns are often illusions.

Here are a few examples of narrative fallacy:

  • History is presented as an inevitable march of a sequence of events, rather than a chaotic mishmash of influences and people. If the past were so easily predictable in hindsight, then why is it so hard to predict the future?
  • Management literature profiles the rise and fall of companies, attributing company growth to key decisions and leadership styles, even the childhood traits of founders. These stories ignore all the other things that didn’t happen that could have caused the company to fail (and that did happen to the many failed companies that aren’t profiled - survivorship bias). Ignoring this,...

Thinking, Fast and Slow Summary Part 3-2: Formulas Beat Intuitions

Humans have to make decisions from complicated datasets frequently. Doctors make diagnoses, social workers decide if foster parents are good, bank lenders measure business risk, and employers have to hire employees.

Unfortunately, humans are also surprisingly bad at making the right prediction. Universally in all studies, algorithms have beaten or matched humans in making accurate predictions. And even when algorithms match human performance, they still win because algorithms are so much cheaper.

Why are humans so bad? Simply put, humans overcomplicate things.

  • They inappropriately weigh factors that are not predictive of performance (like whether they like the person in an interview).
  • They try too hard to be clever, considering complex combinations of features when simply weighted features are sufficient.
  • Their judgment varies moment to moment without them realizing it. System 1 is very susceptible to influences without the conscious mind realizing. The person’s environment, current mood, state of hunger, and recent exposure to information can all influence decisions. Algorithms don’t feel hunger.
    • As an example, radiologists who read the same X-ray twice give different answers 20% of the time.
  • Even when given data from the formula, humans are bad! They feel falsely they can “override” the formula because they see something that’s not accounted for.

Simple algorithms are surprisingly good predictors. Even formulas that put equal weighting on its factors can be as accurate as multiple-regression formulas, since they avoid accidents of sampling. Here are a few examples of simple algorithms that predict surprisingly accurately:

  • How do you predict marital stability? Take the frequency of sex and subtract the frequency of arguments.
  • How do you predict whether newborns are unhealthy and need intervention? A while back, doctors used their (poor) judgment. Instead, in 1952 doctor Virginia Apgar invented the Apgar score, a simple algorithm that takes into account 5 factors, such as skin color and pulse rate. This is...

Thinking, Fast and Slow Summary Part 3-3: The Objective View

We are often better at analyzing external situations (the “outside view”) than our own. When you look inward at yourself (the “inside view”), it’s too tempting to consider yourself exceptional— “the average rules and statistics don’t apply to me!” And even when you do get statistics, it’s easy to discard them, especially when they conflict with your personal impressions of the truth.

In general, when you have information about an individual case, it’s tempting to believe the case is exceptional, and to disregard statistics of the class to which the case belongs.

Here are examples of situations where people ignore base statistics and hope for the exceptional:

  • 90% of drivers state they’re above average drivers. Here they don’t necessarily think about what “average” means statistically—instead, they think about whether the skill is easy for them, then intensity match to where they fit the population.
  • Most people believe they are superior to most others on most desirable traits.
  • When getting consultations, lawyers may refuse to comment on the projected outcome of a case, saying “every case is unique.”
  • Business owners know that only 35% of new businesses survive after 5 years. Despite this, 82% of entrepreneurs put their personal odds of success at 70% or higher, and 33% said their chance of failing was zero!
    • Kahneman tells a story of meeting motel owners who said they bought the motel cheaply because “the previous owners failed to make a go of it.” They seemed blithely indifferent to the circumstances that led the previous owner to fail.
  • CEOs make large, splashy mergers and acquisitions, despite research showing a poor track record of such strategies working.

Planning Fallacy

The planning fallacy is a related phenomenon where you habitually underestimate the amount of time and resources required to finish a project.

When estimating for a project, you tend to give “best case scenario” estimates, rather than confidence ranges. You don’t know what you don’t know about what will happen—the emergencies, loss of...

Thinking, Fast and Slow Summary Part 4: Choices | 1: Prospect Theory

Part 4 of Thinking, Fast and Slow departs from cognitive biases and toward Kahneman’s other major work, Prospect Theory. This covers risk aversion and risk seeking, our inaccurate weighting of probabilities, and sunk cost fallacy.

Prior Work on Utility

How do people make decisions in the face of uncertainty? There’s a rich history spanning centuries of scientists and economists studying this question. Each major development in decision theory revealed exceptions that showed the theory’s weaknesses, then led to new, more nuanced theories.

Expected Utility Theory

Traditional “expected utility theory” asserts that people are rational agents that calculate the utility of each situation and make the optimum choice each time.

If you preferred apples to bananas, would you rather have a 10% chance of winning an apple, or 10% chance of winning a banana? Clearly you’d prefer the former.

Similarly, when taking bets, this model assumes that people calculate the expected value and choose the best option.

This is a simple, elegant theory that by and large works and is still taught in intro economics. But it failed to explain the phenomenon of risk aversion, where in some situations a lower-expected-value choice was preferred.

Consider: Would you rather have an 80% chance of gaining $100 and a 20% chance to win $10, or a certain gain of $80?

The expected value of the former is greater (at $82) but most people choose the latter. This makes no sense in classic utility theory—you should be willing to take a positive expected value gamble every time.

Risk Aversion

To address this, in the 1700s, Bernoulli argued that 1) people dislike risk, and that 2) people evaluate gambles not based on dollar outcomes, but on their psychological values of outcomes, or their utilities.

Bernoulli then argued that utility and wealth had a logarithmic relationship. The difference in happiness between someone with $1,000 and someone with $100 was the same as $100 vs $10. On a linear scale, money has diminishing marginal utility.

This concept of...

Thinking, Fast and Slow Summary Part 4-2: Implications of Prospect Theory

With the foundation of prospect theory in place, we’ll explore a few implications of the model.

Probabilities are Overweighted at the Edges

Consider which is more meaningful to you:

  • Going from 0% chance of winning $1 million to 5% chance
  • Going from 5% chance of winning $1 million to 10% chance

Most likely you felt better about the first than the second. The mere possibility of winning something (that may still be highly unlikely) is overweighted in its importance. (Shortform note: as Jim Carrey’s character said in the film Dumb and Dumber, in response to a woman who gave him a 1 in million shot at being with her: “so you’re telling me there’s a chance!”)

More examples of this effect:

We fantasize about small chances of big gains.

  • Lottery tickets and gambling in general play on this hope.
  • A small sliver of chance to rescue a failing company is given outsized weight.

We obsess about tiny chances of very bad outcomes.

  • The risk of nuclear disasters and natural disasters is overweighted.
  • We worry about our child coming home late at night, though rationally we know there’s little chance anything bad happened.

People are willing to pay disproportionately more to reduce risk entirely.

  • Parents were willing to pay 24% more per bottle of insect spray to reduce a child’s risk of poisoning by 2/3, but they were willing to pay 80% as much to reduce it to 0.
  • People are not willing to pay half price for insurance coverage that covers you only on odd days. Technically, this is a good deal, because there are more odd days in a year than even, so you’re getting more value for your money. But the reduction from 100% risk to 50% risk is far less valuable than the reduction from 50% risk to 0% risk.

We Feel Better About Absolute Certainty

We’ve covered how people feel about small chances. Now consider how you feel about these options on the opposite end of probability:

  • In a surgical procedure, going from 90% success rate to 95% success rate.
  • ...

Thinking, Fast and Slow Summary Part 4-3: Variations on a Theme of Prospect Theory

Indifference Curves and the Endowment Effect

Basic theory suggests that people have indifference curves when relating two dimensions, like salary and number of vacation days. Say that you value one day’s salary at about the same as one vacation day.

Theoretically, you should be willing to trade for any other portion of the indifference curve at any time. So when at the end of the year, your boss says you’re getting a raise, and you have the choice of 5 extra days of vacation or a salary raise equivalent to 5 days of salary, you see them as pretty equivalent.

But say you get presented with another scenario. Your boss presents a new compensation package, saying that you can get 5 extra days of vacation per year, but then have to take a cut of salary equivalent to 5 days of pay. How would you feel about this?

Likely, the feeling of loss aversion kicked in. Even though theoretically you were on your indifference curve, exchanging 5 days of pay for 5 vacation days, you didn’t see this as an immediate exchange.

As with prospect theory, the idea of indifference curves ignores the reference point at which you start. In general, people have inertia to change.

They call this the endowment effect. Before you have something, you might have a certain indifference curve. But once you get it, a new reference point is set, and from this point loss aversion sets in—the utility of a gain is less than a corresponding loss.

Here are a few examples of when people overvalue things once they own it:

  • Sometimes you’re only willing to sell something for higher than you can buy it on the market (like a collector’s item).
  • Famous experiment: half of the subjects were given mugs and were asked to quote prices to sell their mug; the other half, who were not given mugs, were asked to bid for the cups. The buyers had an average bid of $2.87; the cup owners quoted $7.12, over double the buyers’ bid! Note that the cup owners had just received their cups a few minutes earlier.
    • Notably, a third group (Choosers) could either receive a mug or a sum of...

Thinking, Fast and Slow Summary Part 4-4: Broad Framing and Global Thinking

When you evaluate a decision, you’re prone to focus on the individual decision, rather than the big picture of all decisions of that type. A decision that might make sense in isolation can become very costly when repeated many times.

Consider both decision pairs, then decide what you would choose in each:

Pair 1

1) A certain gain of $240.

2) 25% chance of gaining $1000 and 75% chance of nothing.

Pair 2

3) A certain loss of $750.

4) 75% chance of losing $1000 and 25% chance of losing nothing.

As we know already, you likely gravitated to Option 1 and Option 4.

But let’s actually combine those two options, and weigh against the other.

1+4: 75% chance of losing $760 and 25% chance of gaining $240

2+3: 75% chance of losing $750 and 25% chance of gaining $250

Even without calculating these out, 2+3 is clearly superior to 1+4. You have the same chance of losing less money, and the same chance of gaining more money. Yet you didn’t think to combine all unique pairings and combine them with each other!

This is the difference between narrow framing and broad framing. The ideal broad framing is to consider every combination of options to find the optimum. This is obviously more cognitively taxing, so instead you use the narrow heuristic—what is best for each decision at each point?

An analogy here is to focus on the outcome of a single bet, rather than assembling a portfolio of bets.

Yet each single decision in isolation can be hampered by probability misestimations and inappropriate risk aversion/seeking. When you repeat this single suboptimal decision over and over, you can rack up large costs over time.

Practical examples:

  • If given a highly profitable gamble (e.g. 50-50 to lose 1x or gain 1.5x), you may be tempted to reject the gamble once. But you should gladly play 100 times in a row if given the option, for you are almost certain to come out ahead.
  • In a company, individual project leaders can be risk averse when leading their own project, since their compensation is heavily tied to project success. Yet...

Thinking, Fast and Slow Summary Part 5-1: The Two Selves of Happiness

Part 5 of Thinking, Fast and Slow departs from cognitive biases and mistakes and covers the nature of happiness.

(Shortform note: compared to the previous sections, the concepts in this final portion are more of Kahneman’s recent research interests and are more a work in progress. Therefore, they tend to have less experimental evidence and less finality in their conclusions.)

Happiness is a tricky concept. There is in-the-moment happiness, and there is overall well being. There is happiness we experience, and happiness we remember.

Consider having to get a number of painful shots a day. There is no habituation, so each shot is as painful as the last. Which one represents a more meaningful change?

  • Decreasing from 20 shots to 18 shots
  • Decreasing from 6 shots to 4 shots

You likely thought the latter was far more meaningful, especially since it drives more closely toward zero pain. But Kahneman found this incomprehensible. Two shots is two shots! There is a quantum of pain that is being removed, and the two choices should be evaluated as much closer.

In Kahneman’s view, someone who pays different amounts for the same gain of experienced utility is making a mistake. This thought experiment kicked off Kahneman’s investigation into happiness.

Experiencing Self vs Remembering Self

Kahneman presents two selves:

  • The experiencing self: the person who feels pleasure and pain, moment to moment. This experienced utility would best be assessed by measuring happiness over time, then summing the total happiness felt over time. (In calculus terms, this is integrating the area under the curve.)
  • The remembering self: the person who reflects on past experiences and evaluates it overall.

The remembering self factors heavily in our thinking. After a moment has passed, only the remembering self exists when thinking about our past lives. The remembering self is often the one making future decisions.

But the remembering self evaluates differently from the experiencing self in two critical ways:

  • Peak-end rule: The overall...

Thinking, Fast and Slow Summary Part 5-2: Experienced Well-Being vs Life Evaluations

Measuring Experienced Well-Being

How do you measure well-being? The traditional survey question reads: “All things considered, how satisfied are you with your life as a whole these days?”

Kahneman was suspicious that the remembering self would dominate the question, and that people were terrible at “considering all things.” The question tends to trigger the one thing that gives immense pleasurable (like dating a new person) or pain (like an argument with a co-worker).

To measure experienced well-being, he led a team to develop the Day Reconstruction Method, which prompts people to relive the day in detailed episodes, then to rate the feelings. Following the philosophy of happiness being the “area under the curve,” they conceived of the metric U-index: the percentage of time an individual spends in an unpleasant state.

They reported these findings:

  • There was large inequality in the distribution of pain. 50% of people reported going through a day without an unpleasant episode. But a minority experience considerable emotional distress for much of the day, for instance from illness, misfortune, or personal disposition.
  • Different activities have different U-indices. Morning commute: 29%; childcare: 24%; TV watching: 12%; sex: 5%.
  • The weekend’s U-index is 6% lower than weekdays, possibly because people have more control over putting time into pleasurable activities.
  • Different cultures show different U-indices for the same activities. Compared to American women, French women spend less time with children but enjoy it more, perhaps because of more access to child care. French women also spend the same amount of time eating, but they enjoy it more, possibly because they mentally focus on it rather than mindlessly eat in a rush.
  • Your current mood depends largely on the current situation, not on general factors influencing general satisfaction.
    • Things that affect mood: coworker relations, loud noise, time pressure, a boss hovering around you.
    • Things that do not affect mood: benefits, status, pay.
  • Some activities...

Thinking, Fast and Slow Summary Shortform Exclusive: Checklist of Antidotes

As an easy reference, here’s a checklist of antidotes covering every major bias and heuristic from the book.

Cognitive Biases and Heuristics

  • To block System 1 errors, recognize the signs that you’re in trouble and ask System 2 for reinforcement.
  • Observing errors in others is easier than in yourself. So ask others for review. In this way, organizations can be better than individuals at decision-making.
  • To better regulate your behavior, make critical choices in times of low duress so that System 2 is not taxed.
    • Order food in the morning, not when you’re tired after work or struggling to meet a deadline.
    • Notice when you’re likely to be in times of high duress, and put off big decisions to later. Don’t make big decisions when nervous about others watching.
  • In general, when estimating probability, begin with the baseline probability. Then adjust from this rate based on new data. Do NOT start with your independent guess of probability, since you ignore the data you don’t have.
  • WYSIATI
    • Force yourself to ask: “what evidence am I missing? What evidence would make me change my mind?”
  • Ordering effect
    • Before having a public discussion on a topic, elicit opinions from the group confidentially first. This avoids people from getting their first biased by the first speakers.
  • Reversion to the mean
    • Question whether high performers today are merely outliers who will revert to the mean.
    • When looking at high and low performers, question what fundamental factors are actually correlated with performance. Then, based on these factors, predict which performers will continue and which will revert to the mean.
  • Anchors
    • In negotiations, when someone offers an outrageous anchor, don’t engage with an equally outrageous counteroffer. Instead, threaten to end the negotiation if that number is still on the table.
    • If you’re given one extreme number to adjust from, repeat your reasoning with an extreme value from the other direction. Adjust from there, then average your final two results.
  • ...