When rational thinking is correlated with intelligence the correlation is modest | Hacker News Reader

When rational thinking is correlated with intelligence the correlation is modest

scientificamerican.com

120 pointslkrubner83 comments

The author of the article kindly submitted to open this thread has thought deeply about the title issue of the article and has written a whole book on the subject[1] (and parts of several other books on closely related topics). Other researchers on human intelligence give Keith E. Stanovich, the author of the submitted article, credit for bringing up findings from experimental studies on human thinking that will have be taken into account as cognitive psychology refines its theories on how you and I think. I see several comments in this thread along the lines of "The doesn't ponder" or "the author doesn't make the case" as if this article, restricted by a length limit, is the only writing he has ever produced on the topic. He mentions his longer book in the article. The article submitted today is a brief popular summary of an ongoing research program that the author has thought about as deeply as anyone on earth.

[1] http://yalepress.yale.edu/yupbooks/book.asp?isbn=97803001646...

Well, I personally think the current editorialized title here ("When rational thinking is correlated with intelligence the correlation is modest") is off, but the title of the article itself ("Rational and Irrational Thought: The Thinking that IQ Tests Miss") is more accurate.

The article essentially points more to the disparity between IQ test results and rational thinking ability, which is more a critique of IQ testing. However, the title here points to a disparity between intelligence and rational thinking capacity, which is certainly counterintuitive. So, those who didn't note the difference in titles and/or went in with the idea that the title here was the author's premise may have taken issue with the article.

The title editorializing was likely purposeful in that many people dismiss IQ tests out-of-hand, so the notion that IQ tests aren't predictive of rational-decision making capacity would have been far less remarkable or click-worthy.

I consider myself smart and rational, so I was surprised that I stumbled on many of the test questions. I got this one wrong, though you would think, having been programming computers for 15 years, I surely should have gotten this one right:

Jack is looking at Anne, but Anne is looking at George. Jack is married, but George is not. Is a married person looking at an unmarried person?

A) Yes

B) No

C) Cannot be determined

Oddly enough, I feel that if this had been presented to me as a bit of Java or Ruby code, then I would have gotten this one correct. It's the informality of English, and the assumptions I make when reading English, that tripped me up.

Computer code engages the rational side of my brain in a way that English does not.

And to follow up on a point made long ago by Jeff Atwood, I think this (that IQ and rationality are different) also explains the strange inability of some smart people to learn how to program computers. I'm thinking of the post here:

http://blog.codinghorror.com/separating-programming-sheep-fr...

Jeff Atwood discusses this quote:

"All teachers of programming find that their results display a 'double hump'. It is as if there are two populations: those who can [program], and those who cannot [program], each with its own independent bell curve. Almost all research into programming teaching and learning have concentrated on teaching: change the language, change the application area, use an IDE and work on motivation. None of it works, and the double hump persists."

And also this quote:

"Despite the enormous changes which have taken place since electronic computing was invented in the 1950s, some things remain stubbornly the same. In particular, most people can't learn to program: between 30% and 60% of every university computer science department's intake fail the first programming course. Experienced teachers are weary but never oblivious of this fact; brighteyed beginners who believe that the old ones must have been doing it wrong learn the truth from bitter experience; and so it has been for almost two generations, ever since the subject began in the 1960s."

SPOILER ALERT (ANSWER OF RIDDLE BELOW):

I don't really see where you got stuck. Could you mind explaining me why/how/where you got stuck? I couldn't find a satisfying 'aha!' moment so I just kind of brute forced all the possibilities ..

SPOILER ALERT (ANSWER OF RIDDLE BELOW):

Like most people, my first thought was c. And yet, as I said, I suspect that if this had been presented as a bit of Java or Ruby, I would have been unlikely to make that mistake. I read English with a laziness that is different from how I read computer code.

I got it wrong, not due to lack of logic but due to lack of spending the time to untangle the strange wording. I read it as "Jack is looking at Anne, but Anne is looking at _Jack_." and assumed it was some sort of trick question. If they were really interested in testing logic they would deliver the problem in a clear manor rather than masking it strange wording.

The failure mode isn't getting stuck, it's failing to notice that there's actual work to do. The way you get this one wrong is a thought process like this: "Well, obviously we only know about the first and third links in the chain, and we aren't told anything about the second one. So of course we can't tell whether there's a married person looking at an unmarried one because we have no information about any two directly 'connected' people." (But less explicit -- if you think about it in that much detail you're likely to think of brute-forcing the possibilities.) Or like this: "Well, let's see: A looks at B looks at C, A is married, C not. Does A->B have to be married->unmarried? No; B could be married. Does B->C have to be married->unmarried? No; B could be unmarried. So we're done."

Once it occurs to you that you could just kind of brute force all the possibilities, you're going to get it right. The problem comes when you short-circuit the process.

(And the reason why doing so isn't just stupid is that often this sort of "pruning" of mental effort is key to working things out efficiently. It just happens that in this case the prunability is illusory.)

To expand on this.

At first glance, we don't know what value Anne has. But marital status is a boolean value. Since marital status is boolean, it follows by the Law of the Excluded Middle that Anne is either {married | unmarried}. If we brute force the question by simulating each possible value that Anne can assume (and follow the implications), we arrive at the correct answer.

But imagine a case where we test for flavors of ice cream rather than marital status.

> Jack is looking at Anne, but Anne is looking at George. Jack has chocolate ice cream, but George has vanilla ice cream. Is a person with chocolate ice cream looking at an person with vanilla ice cream?

Can we correctly answer the new question? No, because Anne might have strawberry ice cream. A flavor isn't a boolean value!

Also consider that simulation (especially when exhausting every possibility) expends a lot of energy. Therefore, the default strategy is to use a heuristic. And the heuristic is chosen by pattern-matching the "feel" of the question. Unfortunately, the feel of the original question doesn't trigger a heuristic which takes into account the boolean nature of marital status. So rather than assign Anne a value of {married | unmarried}, we assign Anne a value of {null} and call it a day.

And of course it requires assuming "unmarried" is not distinct from "widowed" and "divorced", which may or may not hold up.

>> And of course it requires assuming "unmarried" is not distinct from "widowed" and "divorced", which may or may not hold up.

Not really. Many people consider those other options as special cases of "married" or "unmarried". In that case, how you match them up is not relevant to the problem because she still fits one or the other. She could also be a cat, which would change the answer.

I totally agree that '[m]any people consider those other options as special cases of "married" or "unmarried"'. It's possible it's most people. However, I assert that enough people consider "unmarried" different from "divorced" or "widowed" that the assumption your audience thinks the other way "may not hold up".

I also think the problem lies in the question. It assumes "married" and "unmarried" are mutual exclusive and span the entire space of possibilities. This does not appear to line up with reality in ways I find compelling. Nations may recognize people as married, but due to laws about common law marriage, differing requirements for marriage (very few people who consider themselves married in America filled out the proper paperwork for Tajikistan, a foolish and impossibly rare oversight, the test seems to assume), and many other X factors, it is completely reasonable to assume that people can be neither married nor unmarried, or that George is not married but both married and unmarried, or many other things.

I think if we got code for this, if it used numbers of booleans, or really any system we are familiar with on computers, we wouldn't make this error -- not due to differences in how we think but differences in the object level discussion we would be having.

Jeff Atwood's blog post centers around a paper that was circulated informally ("The Camel Has Two Humps"), but never published in any journal. It has been retracted by its author, who is vigorously trying to undo the harm he believes it has caused.

http://retractionwatch.com/2014/07/18/the-camel-doesnt-have-...

Wouldn't it be nice if Jeff also took down that blog post (or at least put a big, prominent disclaimer at the top about the retracted research paper. The potential harm of such thinking getting pervasive is real (to the younger generation trying to learn the skills).

Simpler explanation: programming teachers are worse than they think they are.

"Jack is looking at Anne, but Anne is looking at George. Jack is married, but George is not. Is a married person looking at an unmarried person?"

On a side note, in constructive propositional logic, the answer is indeed C (http://en.wikipedia.org/wiki/Intuitionistic_logic). In this type of formalism, you have to provide a witness for every existential, or choice for every disjunction. In particular the law of excluded middle isn't an axiom. In order to prove the above proposition, you would need to be able to prove whether Anne is married, or unmarried.

Was going to point this out as well. For a moment I almost resigned myself to be among the 80%, luckily I found a way to smug myself out of self-perceived mediocrity.

I find baring the law of the excluded middle an unnecessarily restrictive basis for logic; but regardless, isn't this a simple (A \/ ~A) case? I mean, you may not be able to prove that Anne is married or prove that Anne is unmarried, but can't you prove by that Anne is either married or unmarried, by the definition of marriege?

Moreover, I had the idea that Intuitionistic logic is advocated to curb with non-constructive proofs over non-finite sets rather than the simple case you mentioned.

"I find baring the law of the excluded middle an unnecessarily restrictive basis for logic"

I agree that it doesn't seem to make sense in that case. However, there are practical use for constructive proofs, for instance, program extraction. Consider a proposition such that "for all list L, there is a list L' such that L' is sorted and L' is a permutation of L". From a constructive proof of that proposition, you can extract a sorting algorithm. This works basically because the constructive proof has to exhibit a witness. That's the kind of things you can do with the Coq proof assistant.

"you may not be able to prove that Anne is married or prove that Anne is unmarried, but can't you prove by that Anne is either married or unmarried, by the definition of marriege?"

You can't prove it in constructive logic. You can always add an axiom of the form "for all x, x is married or x is not married". However, if you do that, your proof won't be constructive anymore (unless you provide a decision procedure together with your axiom).

I find baring the law of the excluded middle an unnecessarily restrictive basis for logic...

And that attitude combined with a long chain of reasoning will wind up with you accepting that the following are entirely sensible statements.

"I accept that there exist real numbers which cannot ever be specified."

"I accept that there exists a reasonable computer science problem for which there exists a polynomial time algorithm but for which I cannot ever prove any algorithm correct."

If you do accept them as reasonable, it is worth carefully considering what the word "exists" actually means to you, and how someone else might think it should mean something different. IF you come to the conclusion that it should mean something different than what is implied by those statements, then you should take another look at constructivism. :-)

There does exist real numbers that cannot be specified in (finite) space or time...

This is what btilly is talking about. To parahrase: "there exist entities that cannot ever be constructed or named in any Universe like ours"

According to classical mathematics, this is true. I can make the statement even stronger. There are real numbers which not only cannot be specified, but are strongly not compressible. That is, for the first n digits, there is no significantly more efficient way to specify what they are than to just list them.

But we are now talking about the existence of something that cannot be specified or written down in any useful way, shape, or form. What does it actually mean to say that this exists? We've demonstrated that no direct evidence of existence can ever be found. This is pretty much the opposite of what a lot of people would actually want "existence" to mean.

From a classical point of view what is wrong is your notion of existence. But it is perfectly reasonable to instead conclude that it is classical mathematics which has gone off the rails and they are now using words like "existence" in a nonsensical way.

Both points of view are perfectly defensible. In fact it was proven by Goedel that any inconsistency in one leads to an inconsistency in the other. Therefore neither is actually able to claim to have a better grounding in logical consistency than the other. Historically the debate was that classical mathematics is more convenient to work with, while constructivism actually means something more reasonable.

But once the set paradoxes faded, and with no difference in consistency, the simplicity of classical mathematics won.

So what's constructivism useful for anyway? Why change the basis of logic instead of proving non-constructability within the reasonable existing logics? (E.g. prove a real number isn't constructable instead of foregoing it's existence)

There are also cases of interest beyond constructivism: what if a number is non-constructible but can be approximated? Saying it doesn't exist doesn't sound particularly useful.

So what's constructivism useful for anyway?

Not much. The philosophy of math is not noted for being particularly useful. :-)

That said, historically constructivism arose during a period when math was having trouble creating solid foundations. Today we can talk about "the reasonable existing logics". But a century ago you couldn't. Attempts to create such a thing had repeatedly resulted in paradoxes, and an inability to find solid foundations had caused repeated mathematical crises for a century.

In that light, it was reasonable to doubt lines of reasoning that proved the existence of things in infinite sets but provided no way to actually find them. This smelled too much of the kinds of things that had lead to paradox.

Hence the appeal of a philosophy that insisted that things are only known to exist when we have constructions for them. Proofs are only valid when they follow lines of reasoning that can, at least in principle, actually be followed in finite time.

These days this is mostly a historical footnote. But an important one.

There are also cases of interest beyond constructivism: what if a number is non-constructible but can be approximated?

Um, actually a constructivist would say that that number exists.

To understand, you have to ask what a real number is. One definition of a real number is a Cauchy sequence of rational numbers. Which is to say an infinite series of rational numbers that converges. Constructivists make this more precise. Rather than talking about an infinite series of rational numbers, they talk about a algorithm that generates a series of rational numbers. Rather than saying "for every epsilon greater than 0 there exists an N..." they insist that there is an algorithm which can take in epsilon and spit out N. Combining these two, from a valid specification of a real number, you can generate an algorithm that given an epsilon > 0 can generate a rational number within epsilon of the real number. And an algorithm that does that can easily be turned into a Cauchy sequence.

Thus to a constructivist, being able to approximate a number arbitrarily well is perfectly fine as a definition of said number. Things get odd in that you can have two different constructions of real numbers but you don't always know whether they are specifying the same number. But you quickly get used to this.

Saying it doesn't exist doesn't sound particularly useful.

Let me turn that around.

What is the use of asserting the existence of things that you cannot construct? What are you going to do with something that you cannot find in an infinite haystack? A thing that may be impossible to find, even in principle?

Thanks for the reply, really enjoying the discussion. What I had in mind is a particular case I heard of, Chaitin's Omega, which can't even be approximated arbitrarily by a fixed algorithm. But still that number seems useful to study it's properties as a number.

Expanding further, that's an issue I've had with definitions of computability also. Certain problems like the halting problem are extremely interesting but their non-computability sort of turns them dead. Approximating them even to a finite degree and studying them profoundly seems very worthwhile, from a practical and theoretical standpoint.

I favor constructivism, btw. I just don't find this particular example to be very convincing for the cause. Take the square root of 2 -- it's a real number that can never be written down with any finite pencil and paper. Except for tricks that define the problem away like using a sqrt-2 base, I will only be able to manipulate approximations of that number. But I can still hold in my hand a square object and know that the diagonal is sqrt-2 times the length of a side. If you want to make objections about the square not really being square due to physical tolerances, then you could look to quantum and atomic physics, where such numbers show up and are measurable to whatever accuracy our experimental apparatus provides.

You are talking about something very different than I was.

To a constructivist, sqrt(2) is a perfectly reasonable number. I can write down an algorithm to approximate it to any desired degree. I can describe its properties. I can find ways in which it arises naturally. I can compare it to rational numbers. It is quite reasonable to talk about its existence.

What constructivists object to are pure existence proofs. Take the numbers that I was describing. A classical proof of their existence would be to demonstrate that the set of such numbers has measure greater than 0. Therefore they both exist, and there are lots of them. But now we're stuck. What can we say that is sensible about them?

There are lots of them! There are some over here. There are some over there. How about some that start off with 1.2345678910111213...? Nope, we can compress the start of that one too easily. How about some that start off with 3.14159265359...? Nope, we can compress the start of that one too easily. How about ones that start off with 0.45399480855940464218372115598800974602442341777409214271555307732203389646276632569332825355032120233072728272048609801012859425779656167823065863748297...? Hrm, kind of hard to tell, we'd need to enumerate the output of a ton of programs and we might not be able to figure out whether some of them halt so if there are, we can't find out. But there are lots of them! Promise!

This is a fairly strong example demonstrating how existence to a classical mathematician does not agree with what a constructivist means.

What if Anne is a dog and can't be married? It was never established that three PEOPLE were looking at each other.

In that case, Anne is unmarried

However, it is not a person.

You posted this same comment twice, and four hours after I originally suggested Anne might be a dog. Could you not at least have gone for a cat?

I accidentally replied under the OP when I meant to reply under this thread, so I stripped out the part most relevant and put it here.

Dog was the first thing that came to my head that would have a proper noun(Name) that wasn't a person. Didn't see a reason to change it? Interestingly though, I do prefer cats.

That's mincing words, considering the whole prpblem a trick because maybe no one is a person. That resoning also shows that the "next item in a finite sequence" questions can have any equally correct answer. It doesn't hold up to Occam's Razor.

Yes, it all depends on the dichotomy of Married/Not Married. If other states are allowed, then it's not possible to determine

Yes. But "background logic" (which classical first-order logic strives to model) is needed to prove meta theorems about any of the constructive logics.

While the object-level proofs in some constructive/intuitionistic logic might do away with the law of excluded middle, the meta-theoretics proofs (proofs of theorems about the formalism) usually use them quite freely.

My point is that there is no formalism that can ignore LEM at all levels (informal --> meta-theoretical--> object).

The constructive logics are quite useful for things like program generation but I am not sure if we should use them as a model for human reasoning.

Can someone explain this further to me? For instance, I thought that constructive logic merely says that not-not-A is not necessarily equal to A. In other words, truth is more about whether A is "provable" while not-A means "not provable", while it's perfectly possible for it to be unknowable whether something is provable.

However, in constructive logic isn't it still fine to restrict a value to one of only two values? For instance, it's still valid to say that (for the sake of argument) someone can only be married or unmarried, in constructive logic, right? So therefore, of someone isn't married, they are unmarried by definition - because we're not talking about values of truth/provability. Right?

So in other words, if we agree that someone can only be married or unmarried (which I take as an assumed premise in this problem), wouldn't the answer still be A in constructive/intuitionist logic?

I'm no expert, but as I understand induction is also a form of this kind of "case reasoning", so to get anything much done you will have to introduce some extra axioms, even if you omit the excluded middle.

Deduction not induction, but yes.

Yes, if you assume (M1(p) or not M1(p)) for a specific M1.

I'm surprised the author doesn't ponder the following: Perhaps higher IQ means a person has more ability to solve these sorts of problems, if they take the time to exert the effort. That ability doesn't guarantee success in these sorts of questions without applied effort, however.

That would mean IQ is still very, very important. But that it is not the only importans factor in solving problems well (eg rationally).

To me, that seems quite plausible. Not sure why the author takes a rather anti-IQ slant as opposed for promoting supplemental measures.

I'd like to split the difference and say that sure, IQ means that a person does have more ability to solve these sorts of problems -- but the fact that it doesn't actually prompt you to apply it is a serious deficit for IQ, and an under-recognized problem.

Smart people get lots of things badly wrong lots of the time. The fact that they're smart can make others (and perhaps themselves, too) assume that they won't be badly wrong, and thus broaden the error.

This

It's one thing to solve an informal problem in a magazine, another one to sit and think about a certain problem, especially when you have hints about weather your guess is right or wrong or previous knowledge (the viral disease test is basically Bayes' Theorem)

Exactly. It's rational to look at these "trick" questions, give the obvious answer, and quickly move on. What do you gain by sweating over them? It's not like they came up on the LSAT and correctly determining the answer determines your future. Flexing your IQ uses a lot of energy... literally.

I get really defensive when my intelligence is on the line. I read this entire article with a combination of pretend aloofness and childlike fear of getting gotten by the gotcha-questions. Apparently, my ego plays into my dysrationalia (for better or worse).

Being defensive is very crippling. It's hard to concentrate, you may feel angry, inpatient or suddenly tired and mind wanders away. Reading speed and comprehension slows down.

On the other hand, being aware and honest of your emotions and how they affect your thinking is important critical thinking skill. When you can realize your defensive attitude you can usually overcome it with effort.

If I remember correctly, high IQ people have less cognitive biases that affect formal reasoning, but high IQ don't reduce biases that distort the ability to question one’s own judgment. Ability to reflect own decisions is correlated with dispositions like "active open-mindedness."

The research of Dr. Carol Dweck is interesting here [1]. It says that students who were praised for "being smart" developed this ego/defensiveness problem and subsequently began to underperform. While students praised for working hard did not, and ended up outperforming the "smart" group. The amazing thing about the research, to me, is that this effect was very prominent and quick, and it happened based on only one sentence being told to the students on one occasion. It did not take years of parents telling their child, "you are so smart". It took one teacher telling the child once, "you did so well, you must have worked hard", for the effect to take place.

[1] http://nymag.com/news/features/27840/

I'm going to assume that IQ does measure general intelligence, since it is highly correlated with all sorts of positive life outcomes that seem to be associated with general intelligence.

Given this, I'm more inclined to draw the conclusion that rationality is not central to intelligence. Most real life problems don't reduce to a logically complex puzzle. As my kung fu teacher once said "fighting isn't like a game of chess". In general, real life problems involve a mix of hard logical constraints, and soft constraints that cannot be reasoned about, but rather rely on intuition.

"shallow processing can lead physicians to choose less effective medical treatments."

But shallow processing can lead ER physicians to choose the correct medical treatment when time is of the essence.

These type of multiple-choice problems annoy me, admittedly because I'm bad at them, but also because you don't always want to over analyze the problem and waste valuable hours when a solution that 'will do the job' is a few seconds of mental processing away.

I don't think anyone disagrees (even the author) that this kind of "shortcut" reasoning is valuable. People want to spend as little energy as possible, and that makes a lot of sense! I'm not a biologist or a scientist, but this probably makes sense from an evolutionary perspective.

The problem is when this kind of thinking leads to terrible results. One example would be thinking the effectiveness of a medical treatment is pretty good, when it's actually worse than not administering the treatment. And doctors are guilty of this, which in this case has serious consequences and is not a multiple-choice riddle!

The question is how to tell the difference ahead of time - when is a heuristic approach going to lead to a "good enough" answer versus a "counterintuitively damaging" answer. It would suck if the only way to know for sure is to avoid shortcut reasoning in all cases.

See also: Do Rational People Exist? http://srconstantin.wordpress.com/2014/06/09/do-rationalists...

A lot of this says, to me, that the human brain doesn't do well with concepts like statistics and probabilities. I think the far more interesting questions are why we do irrational things in the face of clear and overwhelming evidence. For instance, the knowledge of how to become physically fit and wealthy is public knowledge and freely available to anyone with internet access. I know I shouldn't eat that giant piece of cake right before I go to bed, but I do anyway, every night. I think the answer is, we are ruled by our chemical makeup, much more than any of us care to admit. The research discussed in "The Power of Habit" is the most interesting as far as ways to overcome our dependence on chemical/hormonal states ruling our lives.

Most people I know probably think I'm smart enough, yet I still didn't get the right answers to all the test questions posed in the article. I started to wonder how come.

One question giving me no trouble at all was item 5. It was obvious to me that the problem was one I often encounter, deciphering the ratios is a skill developed reading countless medical research reports. After a while it seems natural, it's about something salient in my universe.

I'm guessing the item about marital status of individuals, one looking at another, doesn't align with the salience of usual experience. In normal life we have to be sensitive to the context, and in the question that context was deliberately stripped away. It reduces to a purely abstract notion whereas people, even intelligent ones, really want to know "is she married or not". That information is lacking hence the "C" response.

Some of the other questions are kind of tricky unless one is familiar with the statistical domain they arrive from. Sure there's a logic involved, but learning how to think in those logical terms is unlikely to occur unless it's germane to the individual.

If someone I cared about tested positive for the disease (#4), you bet I'd have good reason to figure out the odds. Otherwise, probably not. The motivational aspect is highly relevant and in many cases the "missing link" between intelligence and worse than expected performance.

I got all of them right except 6. My initial response was that you had to turn over all the cards. Thinking about why I chose this option I realised that I was not trusting the rule that a card must have a number on one side if it has a letter on the other side. My experience has made me highly suspicious that the model I have been given by someone is correct so I always like to test the assumptions of the model if possible.

Indeed.

This question indicates that we should trust one part of what is written, and only question another part. But both parts are written in the same voice, by the same authority.

I like your answer. I got the question 'correct' (in part because I've seen the exact same question years ago) but I wonder if the results teach us more about the questions than the answers.

In many human psychology experiments where humans are said to behave irrationally, I often think they are really behaving rationally when the context of the situation is taken into account.

I had it wrong too, but mostly because I had an issue with the 8 card: IMHO it has to be turned over as it might have e.g. a 'V' on the other side, and that side also belongs to the card, but the answer explicitly states 8 isn't needed because apparently the other side isn't part of the card so the rule only applies to the sides you initially see. A bit of a trick question IMHO.

You got that wrong.

The rule is "vowel on one side implies even number on the other side". The only cards that don't follow this rule are the ones that have a vowel and an odd number. A card with an 8 will never disprove the rule regardless of the letter on the other side, that's why you don't need to turn it.

Yeah I got that now. I feel pretty stupid after I found that out... :)

These books that are kind of pop-science descriptions of concepts like dysrationality are great, but what I'd love is a set of workooks (at various levels, e.g. my kids, adult learners like me, etc.) to systematically teach these skills.

Without just being a list of brain teasers where you learn to recognize each trick.

Do books like this exist? I'd love recommendations.

CFAR (the Center for Applied Rationality)'s goal is to create workshops that effectively train people to be rational.

Probably not what you want, though. They're pricey and not targeted at kids.

Only if you define intelligence as IQ is this true. It's more accurate to say that scoring high on IQ tests doesn't correlate very strongly with reasoning skills.

Or, more interestingly, that scores on IQ tests correlate far more with factors like parental incomes or pleasure reading during childhood than they do with reasoning skills.

I think the Jack / Anne / George problem was not fully formed if they want someone to come to the conclusion that 'A' is the answer.

For example, in the real world there are more "states" a person can be in than "married" or "not married". For example someone can be "separated", which IMO is not fully in the "married" or "not married" states.

To formulate the question properly it needs to say "in this make believe world, all people are either married or not married". This leads me down the path to conclude 'A' is the answer. Not without this extra bit of context.

Datashovel, man, I don't understand how being "separated" could be considered distinct from being married- it's a marital status. It's likely that they're not looking for that variable- 'married' is a binary value, why would one need to clarify?

I don't claim to know all there is to know about marriage, or all the different laws / cultures of marriage in the world throughout the entire history of human civilization. But I guess I wasn't prepared to assume that all possible situations in the world (regardless of culture, continent, country, religion, etc...) can be boiled down precisely to either "married" or "not married" states of being.

I guess I would ask: Why not include the extra context to the question?

I find it odd that there are several people who were tripped up by this, and then seem exclusively to concentrate on the whole "married/unmarried" thing.

The answer, of course, is C, because Anne might be Jack's pet dog or something and therefore isn't a person at all. :-)

Great point :)

Yes, that the married state is a boolean, is not immediately clear. However, I don't think that's why people don't consider the tree. For example, pose the same question in the form male/female.

1. Jack is looking at Anne, but Anne is looking at Eve. Jack is male, and Eve is female. Is a male looking at a female?

Although this question might even be less "boolean" than the married/unmarried one, my hypothesis is that more people will say A.

And, just a nice coincidence that my name happens to be Anne. :-)

Most IQ tests have various scales and subscales, verbal, non-verbal, speed, memory etc... Aside from the few subtests that measure fund of knowledge, speed of reasoning, pattern extraction/recognition and ability to abstract problems define (loosely) what is being measured as "intelligence." The "G" Factor. Note that none of these categories strictly implies rational thinking. You can even make a point that scoring high on any of these categories will make you less likely to make rational decisions.

<rant>It seems there's a lot of parallels to be made between IQ tests and programming language micro-benchmarks. Also, they seem to be equally useful: that is, not very except at the extremes. The corollary is that if you didn't pick your programming language based on a micro-benchmark, you probably shouldn't give too much weight to IQ when hiring.

The one with the probability of the disease makes me always feel like a robot. Okay, so I know, I've to check for the prior. (Bayesian) Statistics 101.

The one with 'jack looking at anne' I glimsed already that 80% chose C, so I spend just a bit more time to consider going through the entire decision tree. It again makes me feel like a robot. :-)

There is something very mechanical to these types of questions and I would wholeheartedly agree that this probably is a dimension that is not correlated with intelligence.

The author doesn't make the case for why these types of reasoning skills are important, or why IQ is important for that matter. This isn't the first time that I've encountered brain teasers, and I know that I dislike them because I generally suck at them. The author characterizes this as being a "cognitive miser" (I also hate it when people make up terms), or not putting forth the effort to thoroughly think through the problem and just going with the easy answer. I can accept that this is what's happening, but what the hell does it mean?

Should we be hiring people who are good at brain teasers? Does that kind of thinking indicate the ability to ship products, come up with novel solutions or understand things in a way that leads to significant increases to the bottom line? In my experience, everybody is some kind of stupid. Even puzzle masters occasionally lose their keys or say embarrassing things at parties. Intelligence and the lack there of is so fractured and interconnected that it doesn't seem like any kind of test is going to be able to measure it accurately, or even nail down what 'it' is.

Some of the smartest people I know were good at questions like this, or maybe I just think they're smart because they're so good at these questions. Big companies love to use them during interviews and there's probably a reason for that. It's definitely a skill I would like to improve myself, but I still don't fully understand the value. Life isn't a series of puzzles, it's a series of problems punctuated by sheer randomness. Being able to take this or any kind of test can't be highly correlated with "success" by most measures, and certainly not in the ways that matter to most organizations.

The article mentions one context where this kind of proper reasoning is highly important: doctors assessing the value of treatments for their patients, the usefulness of screening tests and the likely menaing of test results.

The point is that these are not just brain teasers, these are just distilled examples of problems with which the brain often chooses a short-cut algorithm.

The implication is that analogues of these situations happen all the time, and your ability to apply rational thought is proportional to the likelihood of achieving desirable outcomes.

"(I also hate it when people make up terms)"

The term is from 1991, and is really only a weak specialization of what the words "cognitive" and "miser" mean when you combine them.

http://en.wikipedia.org/wiki/Cognitive_miser

I think the key insight is that there are whole classes of problems where the answer your brain is likely to come up with by default is just flat out wrong. That way when you encounter them in real life you know to break out the mathematics, or sit down with pen and paper, rather than just blithely assuming the "obvious" answer is the right one. (The article has a long list of real life situations where you can run into this kind of thing in para 2.)

Should we be using this as an interview selection mechanism? No, there are more useful and directly applicable ways to see how good they are at the job (and the article doesn't suggest it as an interview filter). Is the ability to solve brainteasers and puzzle questions useful in itself? No, it's pretty useless (and a lot of it is simply having seen the question before). Is knowing how the brain works and what its failure modes are (a) an worthwhile subject of scientific investigation and (b) potentially helpful at a personal level for avoiding mistakes? Yes and yes.

The brain teasers are merely examples.

The real point is that "intelligent" people make these mistakes when it matters, e.g. when deciding on a medical treatment, or where to spend money and time.

> The author characterizes this as being a "cognitive miser" (I also hate it when people make up terms)

?? Where else would terms come from?

I don't understand no. 3.

People are being self interested. Why is that irrational?

I think the idea is that they don't see the bias. And while hurting Germany's export business isn't a problem (from the pov of self-interest), you can imagine other situations where it would be a problem.

To stick with example, say they really do believe these dangerous cars should be banned on American streets (and doing so really would save lives), but when they find out the car is their beloved Ford Explorer, they rationalize their caution away. Now the self-interest is getting them to make dangerous decisions, and they don't even realize that's what's happening.

Rational thinking isn't always linked to intelligence( as problem solving intelligence ) at least in my opinion . I know this might sound irational but oh well.

What if Anne is a dog and can't be married? It was never established that three PEOPLE were looking at each other.

This is what we referred to in school as a "trick question". Very rarely present on standardized testing. We were mostly trained to zero in on the lack of information and then choose "C". The factors that play into who choose what are like more numerous than the author has considered; or at least presented.

This is why society develops bigger and bigger problems at scale: A few clever powerful actors can easily manipulate the masses by virtue of these built-in logical failings of most people, often to achieve their own evil or immoral goals.

Reply on news.ycombinator.com