Why I looked at this book
In this book a Nobel prize winner looks at the way we think and how we can change it. It looks interesting, and I'm hoping for something that really does suggest ways to change how I think and not just amusing stories.Is it necessary to stop yourself from time to time and think more carefully about a problem. I'm also interested in expert decisions. What is needed to become an expert? Is learning a subject sufficient, or do you need plenty of hands-on experience too?Thinking, Fast and Slow
Re: Thinking, Fast and Slow
First Impressions
The book starts by introducing us to system 1 and system 2 thinking. System 1 is intuitive thinking, while system 2 is more considered. So far it looks entertaining, and what Kahneman is saying makes a lot of sense. Will the book change the way I think - well I'll have to see about that.Re: Thinking, Fast and Slow
In the book How to Take a Chance we're asked to imagine a group betting on the throw of a die. It has come up 4 twice in a row. One man believes in lucky streaks and would bet on it coning up 4 again. Another believes in the law of averages and would bet against it coming up 4 again. The first belief is known as the Hot-Hand fallacy, the second as the Gambler's fallacy. In Thinking Fast and Slow Kahneman spend some time in warning us away from the Hot Hand fallacy, but my feeling is he then falls for the Gambler's Fallacy.
We are asked to predict the GPA of Julie a senior in a state university, given that she read fluently at the age of 4. So you take the level of 'smartness' to be able to read fluently at 4 and correlate this level of smartness to a GPA. No, says Kahneman, that is just intensity matching. It doesn't take into account regression to the mean. For example in a two day golf tournament those who score highest on the first day are likely to do less well on the second,
There are a lot of problems with this. Firstly you have to accept that you are really, really being asked to predict Julie's GPA, not just do intensity matching. But prediction is hard (especially about the future). This isn't actually a prediction about the future, as then we'd have to take into account that Julie might be killed in a car crash at age 10, or move to a country which doesn't calculate GPA's. But we certainly would need to adjust our estimate based on the knowledge that Julie has gone to a state university rather than getting a job or getting into an Ivy league university. Kahnemann doesn't mention this. It would seem that the best interpretation of this isn't that we should go into all of these details, but rather just do intensity matching.
So to the second problem. Suppose we accept that we're really being asked to predict Julie's GPA rather than just do intensity matching. This isn't like the golf tournament where there's a lot of day to day variation for each player. Being able to read fluently is a fairly robust measure of smartness. It's not like Julie would forget how to read if tested the next day. One would hope that the GPA measurement is robust as well. It seems that what the regression to the mean is saying is that because Julie did so much better than most in the first four years of her life, she is likely to do less well than others in the next fourteen. This looks like the gambler's fallacy to me. Indeed it seems more likely those who start of ahead are likely to benefit more from their years at school than others. To get regression to the mean you need to match the distribution of reading fluency at 4 to the distribution of GPAs in a wholly unintuitive way.
Thirdly, even if we accept that regression to the mean takes place, who's to say that respondents don't take it into account in their predictions? Remember the matching of distributions is likely to highly unintuitive. It's not as if Kahneman exhibits any real data.
This third objection is the real problem with this book. Supposedly, we readers muddle along with system 1 thinking, while Kahneman being smarter, using system 2 thinking comes to the correct answer. This looks like him being arrogant, but at first I thought, no it isn't, rather it'steaching the readers something useful. But then I began to realise that of course Kahneman gets to the 'correct' answer because he decides what it is. In a related example he tells of how he asked for predictions of how well candidates from an Israeli Defence Forces unit would do in officer school. Predictably the distribution of the prediction matched the distribution of how well they had performed up to that point. "Intenity Matching!" says Kahneman. But this shows how weird prediction using regression to the mean is: we have a good idea of how the distribution of grades will turn out, but in our prediction we have to try not to match that distribution. The amazing thing about this example, though, is that presumably the data exists to show how well the predictions actually turned out. Do find out what it was? No we don't! - it seems we should accept what Kahneman says as more convincing than real data.
Probably my favourite example of the faults of this book is a variant of a well known puzzle.
My answer to this isn't 47, it's that the claims in the question are ridiculous. Even starting with a small fragment of a lily pad, after doubling for 48 days it would cover a vast ocean, not a lake. It's no wonder that people try to answer a question that makes sense and say 24 days.
Then there's the case of Tom W. A psychologist (using tests of uncertain validity) gives a personality sketch of Tom, and it's paint him as pretty nerdy. We're asked to predict which graduate field Tom is most likely to specialise in (out of a choice of 9). Most people choose Computer Science. No says Kahnemann, what you should do is discard the results of the sketch and simply choose the biggest graduate school. But this is nonsense. The sketch gives you information and the question is how much should you discard when told of it's uncertain validity. My guess is that the sketch is pretty accurate - the fact that the tests aren't fully up to scratch is unlikely to mean that Tom isn't actually pretty nerdy, and given that, it seems most likely that he will be studying computer science. Again, I realised that the whole question was based on a made up story of which Kahneman decided the 'correct' result.
There is one case that's based on real data, though. Kahneman analysed the performance of a group of investment advisers working for a firm, each of whose bonus was detrmined by their performance. He found that they was no consistent winner over the eight years of data. These people thought they were competent professionals but the firm was rewarding luck as if it were skill. Kahneman expected them to be deeply shocked by his brilliant analysis, but instead they seemed to be to dull even to be bothered about it.
Well, of course they weren't bothered by it. I would think they were very pleased (although maybe they hid this as they didn't want the success to go to Kahneman's head). He seems to imply that they did no better than chimpanzees, but that isn't the case at all -what he showed is that they did no better than each other. Kahneman doesn't seem to understand what professional means. Not a group of prima donnas, but a group with consistent performance who could substitute for each other if necessary. I would think the firm would be horrified if one of the advisers were consistently better than the others, as then all the clients would want that adviser, and they'd be a much smaller firm. Yes the performance rewarded by the bonus might be due to chance, but chance favours the prepared mind. The firm would want the advisers to put in the effort to take advantage of the lucky breaks when they came along. That was what the bonus was for.
Of course there's a lot more to the book than the problem areas that I've looked at above, and what it says makes a lot of sense. But it's densely written, and not easy to refer back to. Also I'd expected a section at the end summarising the advice in the book for how to live your life, but it didn't really happen. So it's the problem areas that I remember, and I would say that if you want to find out about the important results of behavioural economics it's probably better to look elsewhere.
We are asked to predict the GPA of Julie a senior in a state university, given that she read fluently at the age of 4. So you take the level of 'smartness' to be able to read fluently at 4 and correlate this level of smartness to a GPA. No, says Kahneman, that is just intensity matching. It doesn't take into account regression to the mean. For example in a two day golf tournament those who score highest on the first day are likely to do less well on the second,
There are a lot of problems with this. Firstly you have to accept that you are really, really being asked to predict Julie's GPA, not just do intensity matching. But prediction is hard (especially about the future). This isn't actually a prediction about the future, as then we'd have to take into account that Julie might be killed in a car crash at age 10, or move to a country which doesn't calculate GPA's. But we certainly would need to adjust our estimate based on the knowledge that Julie has gone to a state university rather than getting a job or getting into an Ivy league university. Kahnemann doesn't mention this. It would seem that the best interpretation of this isn't that we should go into all of these details, but rather just do intensity matching.
So to the second problem. Suppose we accept that we're really being asked to predict Julie's GPA rather than just do intensity matching. This isn't like the golf tournament where there's a lot of day to day variation for each player. Being able to read fluently is a fairly robust measure of smartness. It's not like Julie would forget how to read if tested the next day. One would hope that the GPA measurement is robust as well. It seems that what the regression to the mean is saying is that because Julie did so much better than most in the first four years of her life, she is likely to do less well than others in the next fourteen. This looks like the gambler's fallacy to me. Indeed it seems more likely those who start of ahead are likely to benefit more from their years at school than others. To get regression to the mean you need to match the distribution of reading fluency at 4 to the distribution of GPAs in a wholly unintuitive way.
Thirdly, even if we accept that regression to the mean takes place, who's to say that respondents don't take it into account in their predictions? Remember the matching of distributions is likely to highly unintuitive. It's not as if Kahneman exhibits any real data.
This third objection is the real problem with this book. Supposedly, we readers muddle along with system 1 thinking, while Kahneman being smarter, using system 2 thinking comes to the correct answer. This looks like him being arrogant, but at first I thought, no it isn't, rather it'steaching the readers something useful. But then I began to realise that of course Kahneman gets to the 'correct' answer because he decides what it is. In a related example he tells of how he asked for predictions of how well candidates from an Israeli Defence Forces unit would do in officer school. Predictably the distribution of the prediction matched the distribution of how well they had performed up to that point. "Intenity Matching!" says Kahneman. But this shows how weird prediction using regression to the mean is: we have a good idea of how the distribution of grades will turn out, but in our prediction we have to try not to match that distribution. The amazing thing about this example, though, is that presumably the data exists to show how well the predictions actually turned out. Do find out what it was? No we don't! - it seems we should accept what Kahneman says as more convincing than real data.
Probably my favourite example of the faults of this book is a variant of a well known puzzle.
In a lake there is a patch of lily pads. Every day the patch doubles in size. If it takes 48 days for the patch to cover the entire lake, long does it take for the patch to cover half of the lake?
My answer to this isn't 47, it's that the claims in the question are ridiculous. Even starting with a small fragment of a lily pad, after doubling for 48 days it would cover a vast ocean, not a lake. It's no wonder that people try to answer a question that makes sense and say 24 days.
Then there's the case of Tom W. A psychologist (using tests of uncertain validity) gives a personality sketch of Tom, and it's paint him as pretty nerdy. We're asked to predict which graduate field Tom is most likely to specialise in (out of a choice of 9). Most people choose Computer Science. No says Kahnemann, what you should do is discard the results of the sketch and simply choose the biggest graduate school. But this is nonsense. The sketch gives you information and the question is how much should you discard when told of it's uncertain validity. My guess is that the sketch is pretty accurate - the fact that the tests aren't fully up to scratch is unlikely to mean that Tom isn't actually pretty nerdy, and given that, it seems most likely that he will be studying computer science. Again, I realised that the whole question was based on a made up story of which Kahneman decided the 'correct' result.
There is one case that's based on real data, though. Kahneman analysed the performance of a group of investment advisers working for a firm, each of whose bonus was detrmined by their performance. He found that they was no consistent winner over the eight years of data. These people thought they were competent professionals but the firm was rewarding luck as if it were skill. Kahneman expected them to be deeply shocked by his brilliant analysis, but instead they seemed to be to dull even to be bothered about it.
Well, of course they weren't bothered by it. I would think they were very pleased (although maybe they hid this as they didn't want the success to go to Kahneman's head). He seems to imply that they did no better than chimpanzees, but that isn't the case at all -what he showed is that they did no better than each other. Kahneman doesn't seem to understand what professional means. Not a group of prima donnas, but a group with consistent performance who could substitute for each other if necessary. I would think the firm would be horrified if one of the advisers were consistently better than the others, as then all the clients would want that adviser, and they'd be a much smaller firm. Yes the performance rewarded by the bonus might be due to chance, but chance favours the prepared mind. The firm would want the advisers to put in the effort to take advantage of the lucky breaks when they came along. That was what the bonus was for.
Of course there's a lot more to the book than the problem areas that I've looked at above, and what it says makes a lot of sense. But it's densely written, and not easy to refer back to. Also I'd expected a section at the end summarising the advice in the book for how to live your life, but it didn't really happen. So it's the problem areas that I remember, and I would say that if you want to find out about the important results of behavioural economics it's probably better to look elsewhere.
Who is online
Users browsing this forum: Alexjelm, Andreasbqc, Andreasyez and 8 guests
