A machine for jumping to conclusions

Drawing early conclusions is effective if these conclusions are likely to be correct, if the cost of an error is low, and if it saves time and thought. Drawing early conclusions is dangerous when the situation is unusual and the stakes are high. In these circumstances, the risk of error is high and can be prevented by an intervention of System 2.

Context is important in defining the interpretation of each element. The form is ambiguous but one jumps to conclusions directly without even being aware of the ambiguity that has been resolved or avoided. System 1 deals with this.

"Ann approached the bank"

In our heads, we imagine Anne walking quietly towards the bank. We imagine a large building with safes, counters, etc. But again, it's a question of context.

If the previous sentence were: "They were floating gently down the river". We would have imagined a very different scene. Because as a result, the term "bank" is no longer associated with the building, with money, but with the river.

System 1 automatically generated a context, the one that seemed the most plausible and natural. It literally made a bet by creating the most likely context, based on our experiences.

What enters into System 1's creation of this context is current experience and recent events. The choice of context was made without even being aware of it. System 1 keeps no trace of the other possibilities, so one never has the impression of making this choice. It makes a choice without doubt. Doubt and uncertainty belong to system 2.

A Bias to Believe and Confirm

For Daniel Gilbert, in "How mental systems believe", when we hear or understand a new idea, we first try to believe it. We imagine this idea in our head. Then there is a verification process. You must know what the idea means, whether it is true or logical. And that's when we then decide whether to believe it definitively or reject it.

This initial acceptance of the idea, this willingness to believe in it, is what System 1 does. We try to construct a logical idea of the thing in our mind. Even if the idea in question is totally illogical. For example, I'll use Gilbert's example, if I tell you that goldfish eat candy. You're automatically going to gather what you know about candy and goldfish, combine that knowledge, and you're going to deduce that it's impossible. So, you're going to decide to reject that idea. It's this verification process, which Gilbert calls unbelieving, that involves system 2.

He had tests done. He showed participants sentences that were totally false, such as "Lizards love to play Sudoku", and asked them to say whether it was true or false. Of course, no one was wrong. He then repeated the test, asking the subjects to memorize a sequence of numbers at the same time. With their System 2 busy, the participants had difficulty answering the question correctly.

The moral is that when system 2 is busy on other tasks, you believe anything. System 1 is biased and tends to believe absolutely everything. System 2 distinguishes right from wrong but is sometimes busy or saturated. It is even lazy. That's why we tend to be more receptive to advertising when we are tired or depressed.

There is also what the author calls "confirmation bias". For example, when you ask the question "Is Sam friendly? ». You will not get the same answer as if you had asked "Is Sam unfriendly? »

Exaggerated emotional coherence (halo effect)

If you like the president's politics, you'll also like his voice, his appearance, etc. And vice versa. We tend to like or hate everything about someone we like or hate, even things we haven't even seen. It's the halo effect. It influences our perception of people and the world.

For example, who is the most likable person? Probably Alan. And yet they have the same traits. It's the first term on the list that changes the meaning of the following and makes us tend to like Alan more than Ben.

To remedy this halo effect, the author recommends decorrelating the error. For example, James Surowiecki did an experiment. He put several subjects in front of glass jars filled with coins and asked them to estimate the number of coins. Individually, the results were very bad. Some were far too much above the count, others below. But by averaging the results, the result was very close to the count.

So the moral of the experiment is that when you're looking for information, you have to draw on as many sources as possible to arrive at accurate conclusions. What is important to decorrelate errors is also to isolate the sources. Sources that are in contact will influence each other. For example, witnesses to a crime, if they talk to each other before they are heard, will tend to modify the testimony they would have given, because they are influenced by each other when they witnessed the same events. Or, before each meeting, the author recommends that everyone write down their position on the subject on a piece of paper to avoid being influenced by those who speak.

System 1 only deals with the information it has and does not feel the need for additional information even if it is necessary to understand the situation. This is what the author calls WYSIATI: What you see is all there is. And this WYSIATI creates cognitive biases.

Overconfidence: for example in a judgment. A portion of the jury has access to the testimony of both parties. Another has access to the testimony of only one side. And it doesn't actually bother them at all. They don't feel they're missing any part of the story and make a decision without regret or hesitation. What matters is not that there is a lot of information or that it is of good quality, but only that the story told is coherent. It is then believed by our system 1. And our lazy System 2 does not feel the need to doubt.

Framing effects: Between "the chances of survival are 90%" and "the mortality rate is 10%", people will be more reassured by the 1st turn of phrase.