Cristian Popa's Key Ideas from Thinking, Fast and Slow
by Daniel Kahneman

An answer to a number question can be influenced by previous information (even if it is completely unrelated to it). Using this strategy to influence decisions is called priming.

It acts as an anchoring effect, the human starts from the real answer he/she would give and goes up/down towards the number that was proposed (accidentally or not). This anchoring can be more or less strong (e.g. when the actual answer is 10, the suggested one is 5, and the person answers 8, it had a 20% impact).

• Asking somebody whether he thinks an environment tax of 20% is too high or too low. If it is higher than what he/she had in mind, the answer will still be lower, but closer to 20% than it would have been before. Same for when it is higher.
• Telling a judge how many years ago you came to the US can influence the length of the next prison sentence he gives. This is completely unrelated, but it is a number that the judge has in mind and can start from.

People tend to like or dislike everything about a person, thing, event or idea because it is easier for System 1 to work like that. Careful analysis of good and bad sides is much more mentally taxing.

When answering a question, the brain tends to only focus on the available information, disregarding the one missing.

For example: if you are told somebody is intelligent and strong you automatically think they are a good leader.

This leads people to jump to conclusions.

People can associate intensity across subjects that are completely unrelated.

For example: if crimes could be represented in colors, then murder would be a deeper shade than robbery.

The brain by default prefers using System 1 to answer questions. When the actual question can't be answered this way, it aims to answer a proxy question that can be.

For example: When faced with the question "How much would you donate for a humanitarian cause?", the brain automatically turns it to "How much do I actually care about that cause?" and "How much money am I willing to lose donating?"

Experiments with a low sample (N) are much more likely to have extreme results than ones with a high N. Conversely, experimenting with large samples is more precise than with small samples.

This can make us have a skewed view of reality by jumping to conclusions where evidence is not sufficient.

Due to the halo effect and the fact that System 1 deals in averages, not sums, people tend to view things that have higher overall quality as better than ones with lower overall quality, even though the latter may be, in fact, strictly better.

For example: 16 perfect-condition pieces of cutlery and 8 broken ones will be valued less than only the 16 perfect pieces, if viewed separately.

People change their opinion about something based on the ease that they retrieve arguments and examples for it.

For example: you will think a course you're attending is better if you are asked to find 10 improvements rather than 5, because you will have a harder time finding them all.

This applies in the case of news on very rare events that are suddenly an issue (e.g. terrorism, plane crashes). This leads to "Availability Cascades" where the people's emotions turn to irrational fear over insignificant risks.

Success or failure are typically outliers due to luck and people/things regress to their usual performance (the mean, or the prior).

For example: in a sports competition, an athlete that does much better than the rest on the first day will most likely perform worse on the following ones, because he attained the initial performance in part due to chance.

Intuitively, this is how success is defined:

• success = talent + luck
• great success = a little more talent + a lot of luck

An easy way of making predictions by taking into account uncertainty and the natures of the 2 systems:

• Start with an estimate of the average answer;
• Determine your estimate (performed by System 1);
• Estimate the correlation between your evidence and the target (performed by System 2);
• Use the correlation to move that many percents from the average to the estimate.

For example: to estimate your GPA, start from the average GPA and set your expected GPA. Estimate how confident you are in your prediction. The predicted GPA will be closer to your expectation if your confidence is higher.

Not all domains are equal in nature. Some are more reliant on skill (e.g. surgery), while others are on luck (e.g. stock trading). People can be tricked into thinking that their success is actually based on skill despite the nature of their activity.

Experimentally speaking, intuition is less likely to come to a better conclusion than simple formulas, that are not influenced by human nature.

For example: you will find more success in hiring candidates if you use a pre-determined checklist instead of relying on your interviewer sense.

The importance of intuition increases based on the domain and its dependence on skill.

When it comes to decision-making, there is a distinction between two idealized "species" of people:

• Humans - Beings with flaws and feelings, susceptible to each of the biases presented here
• Econs - Beings that are always able to think rationally, unaffected by the biases presented here

The "econ" term is the short form for "economist", as people that analyze and invest in the financial markets make every decision consciously, with a clear goal in mind.

The degree of "utility" as perceived by people decreases gradually. $1,000 is perceived wildly different for somebody with a net worth of $10,000 vs. $1,000,000. Similarly, an increase from $1,000 to $2,000 is perceived as much better than one from $10,000 to $11,000 despite the equal increase.

This is the utility theory developed by the mathematician Daniel Bernoulli.

This theory is not ideal and cannot model every case. It fails in certain cases, which is why prospect theory was developed.

Prospect theory is an improved version of the previous utility theory.

Utility is instead based on differences from the reference point.

Humans have, by default, some degree of loss aversion, so the absolute utility of a loss will be perceived as greater than one of an equal gain. A survey found that people would refuse a bet where they would gain $100 with a 50% chance and lose $100 otherwise. In fact, they would not accept until the gain is at least twice as big, depending on their degree of loss aversion.

Thus, there are 3 cognitive features in prospect theory:

• Evaluation based on a reference point
• Diminishing sensitivity (borrowed from Bernoulli)
• Loss aversion

The theory does not take into account human feelings of disappointment when failing to win gambles.

People become more unwilling to let go of the things or benefits they already own. They will only sell them for more than they would be willing to buy them for, even if some time ago they were indifferent to them. In short, they grow attached.

Depending on the situation, people can be either risk averse or risk seeking. It is all about the probability of events happening and whether the end result will be positive or negative. For instance, when the chances are very much in favor and the result is positive, people are risk averse and fear the disappointment that luck won't be on their side, so they do anything to maximize their chances to 100%.

The change from 0% to 1% chance (possibility) and from 99% to 100% chance (certainty) are important psychological thresholds.

People tend to overestimate the probability of rare events and then overweight them in their decisions.

An event is more likely to be preferred if its image is more vivid (e.g. receiving $100 in a blue envelope as opposed to only receiving $100).

People (through System 1) are more sensitive to formulations like "12 out of 1,000" and will sometimes think it's more likely than a risk of 2%. Why? The image is more vivid and it is easier for System 1 to think in terms of individual cases as opposed to percentages.

Additionally, people (again, System 1) will tend to be more sensitive to higher values of the denominator while disregarding the numerator: 100 out of 10,000 may be perceived as worse (or better, depending on the context) than 12 out of 1,000. This is called Denominator Neglect.

When faced with a gamble or chance event, before declining it because of other biases (e.g. loss aversion), think not about the single instance where you win or lose, but rather about your whole life as a series of such gambles and whether the total number will lead to consistent gains. Think like a trader! This is called Broad Framing.

Acting differently from the default option has the downside of bringing regret if the decision turned out to be wrong and defaulting was the best course of action.

Example: you will feel better if you are injured at the convenience store you always go to, as opposed to getting injured at a new convenience store because the other one is doing renovations.

You can reach more sensible decisions when you have a broader context.

In real life, judges are asked not to consult more than the case given to them and, thus, lack the broader context, leading to worse decisions regarding penalties.

Given all the previous information, you can guide people to a decision based on the framing of the problem.

Take loss aversion into account - try to disguise the decision as a win (or loss if you prefer them not to take it).

Think about the reference point and exercise thought of different ones (e.g. imagine you did not own X).

Think about the different mental accounts that are not, in fact, separate (e.g. gains on stocks vs. crypto) and imagine they were merged.

Make the default option the one you want and the differences will be huge. For instance, in the case of governmental systems, the percentage of the population that donates organs on death is largely dependent on whether the citizen needs to submit a request to donate or to not donate on death. In 2003 this percentage was close to 100% for Austria and 12% for Germany. The difference lies in what the default option was.

Finally, along with the two systems and the two species, humans also have two selves:

• Experiencing Self - the self that experiences things in the moment;
• Remembering Self - the self that remembers experiences, remembers them, assigns a rating, and likes creating stories.

The remembering self leverages System 1 and, when it comes to experiences, it cares about:

• The peak intensity (be it negative or positive);
• The way they end;
• To some extent, the average experience but not the duration, because System 1 deals with averages, not sums. This is called Duration Neglect.

People can be determined to consider a strictly worse experience as better than another based on the impression of the remembering self.

Example: in an experiment, people had their hands in very cold water for 60 seconds, then in cold water for 60 seconds and 30 seconds in slightly warmer water. When asked which experience they would rather repeat, most people chose the second option despite it being strictly worse.

The previous result makes sense because people tend to answer harder questions like "Which would you rather repeat?" with easier questions like "Which did you like more?"

This way they can be convinced to repeat a worse experience just because it ended slightly better or had a lower peak intensity even though it took much longer. This is also why people would rather go through a long and painful surgery than an extremely painful surgery that takes a few seconds.

The neglect of experiencing self and the appeal of remembering self (even though both selves constitute us) leads people to like stories more than anything else.

This is why we tend to ignore details that do not fit our view of the story, whatever it is: relationship, career, trips, etc.

Predicting the future from current circumstances, heavily influenced by emotion (e.g. how your marriage will go based on your wedding day) is not ideal.

Because of AYSIATI, people will attribute larger weight to things they are thinking about the moment.

People will regress to the mean and the happiness they experience will always turn back to an average value. For most of us, this average value is 7 on a scale from 1 to 10.