Thinking, fast and slow - D. Kahneman

The book is very interesting. Parts I-III, which deal with System 1 and System 2, are very good, Parts IV and V I think are rather forgettable.

Given that le elucubrazioni from the book stem from various experiments, the analysis remains often superficial. From experiments you can tell that the unconscious mind works a certain way, testing why it is not possible - or in any case was deemed out of the scope of the book from the author, apparently.

I am of course in favour of understanding the way the unconscious works, and it seems that there is a general bottom line (i.e. my unconscious regularized my information intake selectively excluding things that do not fit the pattern) among different ways of looking at the brain, which goes: try to be conscious of what you’re doing/why you’re doing it, things are going to improve. This is at least easy to keep in mind. What I am in not in favour of is the practice of saying, like in this book, that some process has been selected by evolution to work in some way and then without missing a beat claim it should be avoided. A discussion is in order on why whatever rule evolution selected is not valid here, and if a logically robust conclusion cannot be reached things should be left as they are.

This is probably more or less what Taleb writes in his books. In a natural settings, the way the body and mind works is already optimal; in an artificial setting, this might or might not be the case. A question the book does not answer (there are two lines thrown in somewhere, and that’s it) is why in the world one should apply statistic reasoning to unique or sparse events. Of course, if the game is deciding how much to bet when flipping a perfect coin five thousand times I should think statistically, if the game is deciding whether I want to move to marry then I definitively shouldn’t. Say that I can determine that statistically taking the decisions of moving is good (meaning: repeated N times with N large leads to a good outcome most of the times), then I move and it does so happen that I am miserable. Is thinking statistically a good idea here? Of course, say that I make some bet on five thousand coin throws and it comes out five thousand heads, then you just call the cops - it is theoretically possible, but nobody is going to believe it anyway, so you’re good.

If I remember correctly the marriage situation is where Taleb suggests the via negativa, i.e. walk out of this kind of situations if the bad outcome is not deemed tolerable/if there is no safety net, independently of any statistical thinking. I would also argue that evolution thought the same way if our unconscious often does not think statistically - otherwise the statistically savy organism would have prevailed. So I think one should tread carefully here. In other points it seems like the behaviour of the unconscious rather has to do with limitation/economy on the computing power, which is a different story. All of this is hard to test and experiment, which is why I think is not discussed, I think it is a pity and that it takes a lot away from the conclusions.

After having read about it I am also not sure I understand the fuss about loss aversion. The typical example is that you normally wouldn’t take a bet that goes like: we flip a coin and if it’s head you win 120€, if it’s tails you give me 100€. This is supposed to be absurd because the expected value is 10 and 10 is bigger than 0. But while you can win up to infinity, you can only lose down to 0, so I would say it is fairly natural that the scales are tipped - the asymmetry is there from the beginning. You can have 0 apples, 10 or 10 millions, but never -7. One should/could take the log to change the half axis to a whole axis, which goes a bit in the direction of utility, but is not quite the same because of the absence of a reference point. Variance is also not mentioned in the book, but I think it should be considered.

I liked the “law of small numbers”:

The law of small numbers asserts that the law of large numbers applies to small numbers as well.

This is of course widely applied. I always liked the fact that a normal theorem gets the label of a LAW, which tells a lot of the way it is perceived and used.