148 of 178 people found the following review helpful
Lots of Truth, Too Much Hype,
Verified Purchase(What's this?)
This review is from: Thinking, Fast and Slow (Hardcover)
I was privileged to have Daniel Kahneman for many years as a member of my research group "MacArthur Network on the Origin and Nature of Norms and Preferences," where I came to appreciate his dedication, intelligence, and friendship. At last Kahneman has written a book for the public, and is already a widely discussed best seller. I am convinced that the contributions of Kahneman and his coauthor Amos Tversky are fundamental and lasting, but I am concerned that his work not be seen as conveying the general message that "people are irrational."
I was motivated to write these hasty comments by the review of the book by Jim Holt in the New York Times Book Review (11/27/2011), p. 16ff. Holt writes "Although Kahneman draws only modest policy implications... others... go much further. [David Brooks, NY Times editorial writer], for example, has argued that Kahneman and Tversky's work illustrates `the limits of social policy'; in particular, the folly of government action to fight joblessness and turn the economy around." Of course, nothing of the sort follows from Kahneman and Tversky's work.
I know Kahneman and Tversky's work (hereafter KT) very well, but I am going through this book slowly, so I will add to my comments as they accumulate.
Psychologists have known for many years that humans make systematic visual mistakes, called "optical illusions." This does not elicit the general pronouncement that humans systematically error in their visual judgments. The same should apply to KW's results: the results are correct, but they should not be sloppily interpreted as saying that people are general illogical and error-prone decision-makers.
First main point: I believe KW's work does not at all suggest that people are poor at making logical inferences. The experiments which might suggest this are generally misinterpreted. A particularly pointed example of this heuristic is the famous Linda the Bank Teller problem, first analyzed in Amos Tversky and Daniel Kahneman, "Extensional versus Intuitive Reasoning: The Conjunction Fallacy in Probability Judgment", Psychological Review 90 (1983):293-315. Subjects are given the following description of a hypothetical person named Linda: "Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice and also participated in antinuclear demonstrations." The subjects were then asked to rank-order eight statements about Linda according to their probabilities. The statements included the following two: "Linda is a bank teller" and Linda is a bank teller and is active in the feminist movement."
More than 80\% of the subjects---graduate and medical school students with statistical training and doctoral students in the decision science program at Stanford University's business school---ranked the second statement as more probable than the first. This seems like a simple logical error because every bank teller feminist is also a bank teller. However, there is another interpretation according to which the subjects are correct in their judgments. Let p and q be properties that every member of a population either has or does not have. The standard definition of "the probability that member x is p" is the fraction of the population for which p is true. But an equally reasonable definition is the probability that x is a member of a random sample of the subset of the population for which p is true.' According to the standard definition, the probability of p and q cannot be greater than the probability of p. But, according to the second, the opposite inequality can hold: x might be more likely to appear in a random sample of individuals who are both p and q than in a random sample of the same size of individuals who are p.
In other words, the probability that a randomly chosen bank teller is Linda is probably much lower than the probability that a randomly chosen feminist bank teller is Linda. Another way of expressing this point is that the probability that a randomly chosen member of the set "is a feminist bank teller" may be Linda is greater than the probability that a randomly chosen member of the set "is a bank teller," is Linda.
I believe my interpretation is by far the more natural. Moreover, why would the experimenters have included information about Linda's college behavior unless it were relevant? This behavior is completely irrelevant given KW's interpretation of probability, but wholly pertinent given a "conditional probability" interpretation. The latter can be colloquially restated as "the conditional probability that an individual is Linda given that she is a feminist bank teller is higher than the conditional probability that an individual is Linda given that she is a bank teller."
Second Main Point: Many of the examples of irrationality given by KW are not in any way irrational. Consider for example the chief investment officer described by Kahneman (p. 12). This man invested tens of millions of dollars in Ford Motor Company stock after having visited and automobile show and having been impressed with the quality of the current offering of Ford vehicles. Kahneman says "I found it remarkable that he had apparently not considered the one question that an economist would call relevant: I Ford stock currently underpriced?" In fact, there is no objective measure of a stock being "underpriced," and no known correlation between a measure of "being underpriced" and subsequent performance on the stock market. Moreover, the executive may not have revealed all of the reasoning involved in his decision, but rather only a "deciding factor" after other considerations had been factored in.
Third Main Point: We have long known that people do not generally act in their own best interest. We have weakness of will, we procrastinate, we punish ourselves for things that are not our fault, we act thoughtlessly and regret our actions, yet repeat them, we become addicted to cigarettes and drugs, we become obese even though we would like to be thin, we pay billions of dollars for self-help books that almost never work. KW have not added much, if anything, to our understanding of this array of bizarre behaviors. Of course, they do not claim otherwise. However, commentators regularly claim that this behavior somehow contradicts the `rational actor model' of economic theory, which it does not in any way. Economic theory explores the implications of human choice behavior without claiming that the choices people make are in some sense prudent or even desirable to the decision-maker (we cannot choose our preferences).
Economic theory is in general supportive of the notion that people should get what they want, but has included the notion of "merit goods" that society values or disvalues for moral or practical reasons that counterindicate consumer sovereignty. For instance, we regulate pharmaceuticals, we prohibit racial discrimination in public places, and we outlaw markets in body parts.
Fourth Main Point: KW are right on target in asserting that people make massive errors in interpreting statistical arguments (e.g., the base rate fallacy, or the interpretation of conditional probabilities). This has nothing to do with "illogicality" or "irrationality," but rather the complexity of the mathematics itself. For instance, KW have shown that physicians routinely fail to understand what the statistical accuracy of lab tests mean---the fact that at test is 95% accurate is compatible with the fact that it is wrong 95% (or any other, depending on the incidence of the condition that is test for) of the time.
The psychologist Gerd Gigerenzer has shown that if conditional probabilities are reinterpreted as frequencies, people have no problem in interpreting their meaning (see the discussion "Risk School" in Nature 461,29, October 2009). Gigerenzer has be promoting the idea that trigonometry be dropped from the high school math sequence (no one uses it except surveyors, physicists, and engineers) and probability theory be added. This sounds like a great idea to me.
Of course, if people do not do well a formal statistical analysis, how are to we defend the rational actor model, which is thoroughly Bayesian and implicitly assumes people are infinitely capable statistical decision-makers? The answer is not to abandon the rational actor model, which in general has had exceptional explanatory power---see my book, The Bounds of Reason (Princeton 2009) and my review of Ken Binmore's Rational Decisions (Economic Journal, February 2010). Rather, I believe the answer lies in replacing the subjective prior assumption of the rational actor model with a broader assumption that individuals make decision within networks of minds that are characterized by distributed cognition, much as social insects, except of course on a much higher level, using language instead of pheromones, with the cultural construction of iconic rather than pheromonic signals. But, that is the subject to be explored in the future. One of the payoffs of KT research is to make it clear how insufficient the standard economic model of decision-making under (radical) uncertainty really is.
Tracked by 7 customers
Sort: Oldest first | Newest first
Showing 1-10 of 26 posts in this discussion
Initial post: Nov 29, 2011 5:19:47 PM PST
Very interesting. Thanks for this.
Posted on Dec 2, 2011 11:10:31 AM PST
Dr. Chuck Chakrapani says:
Thank you for your extensive and considered review.
Like you, I have great respect for K&T's work and, like you, I also feel that they assume what the "correct' answer should be to the questions they pose. "Linda's problem" always bothered me. While K&T are technically correct in their probability assessments, I believe people tend to answer that question from a different perspective. It's akin to asking someone in a pub with 100 patrons, 10 of whom are drunk and belligerent, "Which is more probable -- you are likely to be attacked any of the 100 patrons, or by one of the 10 belligerent drunks?" The "correct" answer is by anyone of the patrons. Yet most people would say by one of the belligerent drunks and it is a logical answer too. You have identified many other instances in which there are alternative explanations like this. In my Amazon review of the book, I also point out possible alternative explanations for some of Kahneman's assumptions. I feel (you'll probably agree with me) that Gerd Gigerenzer's work in this area deserves a wider audience.
I never had the opportunity to meet with Kahneman, although I had the honor once of presenting at a Decision Making conference in London, where Tversky was also presenting. I presume that you will be updating your review as you go along. I look forward to reading it.
In reply to an earlier post on Dec 22, 2011 1:53:32 PM PST
Not sure whether it's Kosher to respond to a comment rather than to the initial review, but I'm going to do so anyway. Actually I guess it is OK, because Amazon is telling me that what I am typing is "Your reply to Dr. Chuck Chakrapani's post."
I think the pub argument actually gives more justification to the vast majority giving the "wrong" answer than does Herbert Gintis's "a random sample of the subset" argument.
It almost becomes an issue of English rather than math. "Any of the 100 patrons" technically means "a particular one of the 100 patrons", rather than "the totality of the 100 patrons, considered individually."
Since the odds of being attacked by a PARTICULAR one of the 100 patrons (let's say Joe over there at the far right table) is lower than the odds of being attacked by a particular one of the 10 drunk & belligerent ones (let's say Linda who's standing on the bar screaming about how she's going to kill anyone with an ATM card), then it is actually quite logical to answer "I am more likely to be attacked by one of the 10".
As I've tried to write this, I've realized that it's actually very hard, in English, to craft a phrase that unequivocally means "the entire group, considered individually but summed together as a total." "Any of the 100" really means "a single person", and "anyone of the 100" isn't grammatical, and "all of the 100" implies that 100 people are coming at you at once. The most clearly descriptive phrase I can think of is the null phrase. If you ask "Which is more probable -- you will be attacked, or you will be attacked by one of the 10 belligerent drunks?" then I suspect the number of people giving the "logical" answer will increase.
Fascinating thought experiment, I really appreciate Dr. C bringing it up. Although I confess that I don't see how it is "akin" to the Linda problem, where the correct answer still seems unassailable to me.
Posted on Dec 29, 2011 1:14:19 PM PST
Marvin G. Scoggin Jr. says:
While I deeply appreciate the analysis and discussion by H. Gintis, I take exception to the statement, "Gigerenzer has be[en] promoting the idea that trigonometry be dropped from the high school math sequence (no one uses it except surveyors, physicists, and engineers) and probability theory be added. This sounds like a great idea to me." Noooo!!! My life would be severely diminished had I not taken that wonderful one-semester course in trigonometry in high school. And how would I have fared in four semesters of college calculus (I was a chemistry major) without trig? No, I don't make formal use of trigonometry very often in my profession today (I'm a teacher), but I perceive the world differently because of it, seeing those amazing mathematical relationships in concrete form--both natural and human-made--all around me. Whether I'm working around the house, tinkering under the hood of my car or truck, or trekking through the wilderness on a hike or expedition, I often find myself either marveling at the uses made of trigonometric relationships or using them myself to solve structural, mechanical, or directional problems. I'll grant you that statistics should be taught (I had my first statistics course in graduate school over 20 years ago--unbelievable to me that it wasn't a prerequisite for physical chemistry 38 years ago), but I subscribe to the "both-and" solution, not "either-or." Other considerations: Not everyone becomes a musician, but I'll wager that anyone who's had formal instruction in music has a deeper appreciation for and enjoyment of music. Same for art, theater, literature, and just about any other pursuit you can think of. Have you seen the HBO movie, TEMPLE GRANDIN? I highly recommend it for a number of reasons but especially, to the point of this discussion, to illustrate what a difference an understanding of trigonometry makes in appreciating how the main character thinks and in making sense of the way the filmmaker portrays the main character's thinking graphically. This dear soul (she is a real person--now a professor at Colorado State University) even makes the point that people like her (she's autistic) are hopelessly lost in abstract math like algebra (and perhaps she would include statistics) but absolutely liberated and even energized by concrete math such as trigonometry. I look forward to reading Daniel Kahneman's book, and I certainly agree that at least a little bit of statistics should be served on everyone's plate, but along with, not at the expense of, trigonometry!
In reply to an earlier post on Dec 29, 2011 11:50:30 PM PST
Last edited by the author on Dec 29, 2011 11:53:09 PM PST
S. Prewitt says:
Re: "I'll grant you that statistics should be taught ... but I subscribe to the "both-and" solution, not "either-or."
Life is easy when there are no trade-offs.
In addition to statistics, I think all high school students should be taught at least one year of accounting and one year of microeconomics. And they should spend a semester studying the history of our country's journey from republic (when founded) toward the dumbocratic cliff. And all high school students should be required to study--to actually read--the Bible as literature.
And here's the best part: since there are no trade-offs, all of these subjects can be added to the current course load!
In reply to an earlier post on Jan 1, 2012 11:57:22 PM PST
Dr. Chuck Chakrapani says:
Yes, it is completely kosher to reply to a post directly.
Thank you for taking the time to deconstruct the probability argument in simple English. I'm not quite sure if I can return the favour, but let me try. The equivalence (in terms of structure, not meaning) between the Linda problem and the pub problem goes like this:
1. All bank tellers = 100 pub patrons
2. Bank tellers active in the feminist movement = 10 belligerent drunk pub patrons
Since in both problems  is a subset of , the "correct" answers are "linda is a bank teller" and "any of the 100 patrons". But it is so only if you think of the questions in exactly the way they are posed, and not in the way most people would interpret them. Hope this makes it clearer why these two problems are similar. If not, i need to work on my communication skills.
Posted on Jan 5, 2012 12:20:31 PM PST
W Lorraine Watkins says:
"I am concerned that his work not be seen as conveying the general message that "'people are irrational.'" ---- Bingo!
Posted on Jan 27, 2012 10:07:54 PM PST
R porter says:
Like you, I find that many of the examples of irrationality given by KW are not in any way irrational. The Linda question is the most prominent but there are many others, as you note. I find very upsetting the almost limitless range of phenomena to which KW seems to think his theory applies. Some are truly ridiculous. Thanks for your measured but pointed critique.
Posted on Feb 5, 2012 9:01:14 AM PST
MMJ fan says:
OK, a test result will be Thus 95% of the time, and 5% of the time it will not be that result. Whether Thus is "right" or "wrong" is another thing entirely. If a test result comes out one way 95% of the time, it cannot be something else 95% of the time.
In reply to an earlier post on Feb 5, 2012 9:30:52 AM PST
W Lorraine Watkins says:
This may be a lot of parlor fun for the "thinkers." It is pretty significant for sick people who are victims of guidelines for treatment that clip off the ends of the curve.