This isn't true at all. They haven't been shown to be biased, merely that blacks score lower. Extensive research has show that IQ tests have equal predictive power for blacks and whites. "Disparate impact" merely means that blacks score lower, not that the tests are biased.
> This is easy to do with IQ tests, just ask lots of questions that require good schooling and education (e.g. word questions).
The black-white gap is as large or larger on tasks that require very little education, e.g. "reverse digit span", which merely requires that you repeat a sequence of numbers in the reverse order from the direction you heard it.
> and has hence probably improved equality.
I've never understood why eliminating one of the few objective measures and making the hiring process entirely subjective was an improvement.
When I think about that question, I think about questions that would be harder for me (white caucasian) than somebody from another "race" or culture. That is easy: for example imagine selecting some kind of image puzzles involving kanjii symbols, or telling apart chinese faces. They are just images, right, so no bias? But it will be much more easy to tell them apart for somebody with a chinese background than for me. If such questions exists that are biased against me, it seems at least very plausible that other questions could be biased against other "races" and cultures. Another example I read about is how different cultures see the drawing of a box (a 3d wireframe). Apparently in western cultures people would see it as a drawing of a 3d box, whereas some poor test takers in Africa just saw a 2d drawing. This was then interpreted as Africans being less intelligent, when really they were just as correct as the people who saw the 3d shape. It just wasn't such a common meme where they lived to draw 3d shapes like that.
1. how many seconds did it take you to think up this refutation of IQ tests? 2. how many psychologist/psychometrician man-years do you estimate go into developing any particular IQ test like the RAPM or BOMAT? 3. what probability do you give that psychometricians have not also thought of your objection? 4. conditional on their thinking of it, what probability do you give that they have done nothing to cope with it? 5. how many seconds would it take you in Google or Wikipedia to find the real answers to 3 & 4? 6. how many seconds did it take you to write your comment?
Over the years I have read a lot of articles on the subject - as did many others, I guess, it is a popular subject after all.
I'd say it took me more than a few seconds to come up with my example.
First of all, the thing with the 2d vs 3d drawing was an actual real world example that I read about. So clearly there are some psychologists/psychometrians who were wrong about the interpretation of their tests.
As for 5: if it only takes a few seconds to find those answers, why don't you do it and give the answers to us?
Edit: I just tried googling "RAPM cultural bias" and didn't get any useful results. You really should provide the links. Of course your tone makes it likely that discussing with you is probably a waste of time anyway :-/
Not really. I ask these questions because I've noticed that astonishingly often, people online will write a definitive refutation of some scientific position or new study, which took them only a few seconds to think of and which is wrong; but why do they do this?
I can think of many possibilities, and I'm never sure what it is: do they think so poorly of the scientists conducting the study (anti-intellectualism)? do they think so highly of their own thoughts (Dunning-Krueger)? are they too stupid to realize any of this (low IQ)? are they too lazy to look up the answers in Wikipedia or Google Scholar? or is it a cynical ploy for getting upvoted comments where they count on their fellows not being knowledgeable enough to call them on it?
Hence, I ask questions to try to isolate what might be the issue and incidentally point out to the more reflective readers why comments like yours are fundamentally bad ideas.
Sometimes the spectacle is just horrifying; on a mailing list with a fairly high level of discussion of cutting-edge research on working memory and dual n-back, someone posted proudly about how they thought IQ tests were biased against Africans, and though I point out that this was obvious to him and me, even more obvious to psychometricians, he never once looked up how the tests might have been defended against or corrected for cultural biases, despite me sometimes explicitly naming the names he could punch into Google or Wikipedia.
I am still waiting for your refutations. But honestly, your initial assumption that every internet commenter is wrong will make it difficult for you to communicate or learn stuff.
Frankly, I fail to see how pointing out possible flaws in a theory can ever be "a bad idea". In my opinion to doubt everything should be the default stance of a scientist.
You mentioned RAPM, I looked it up and it says it was invented 1934 in an attempt to avoid cultural biases - therefore it uses more graphical elements than language. My examples show that even graphical elements can have a cultural bias. So while it is a nice attempt, you can not be sure RAPM succeeded, even if their intentions were good.
The 2d box vs 3d box thing is an actual event from the history of IQ research, where IQ testers apparently got it wrong. So these people are fallible, too. Finally, cultural bias is one of the standard mistakes people (Even if they are scientists) make.
Note that I didn't even claim the IQ tests are flawed. It bores me if people confuse my comments with stuff other people said.
A great example is Jim Crow era voting rules. Black people obviously are not biologically disinclined to be able to read, but in 1900, it was a simple fact of southern American culture that a white person of voting age was less likely to be able to read than a black person of voting age. A black southerner of voting age was usually either a former slave, or the child of a former slave, and slaves did not have access to basic education. So a literacy test was, at that time, a racially biased test.
---
¹ This is unsurprising when you realize, for example, that the combined population we put in the single box of "black" is more genetically diverse than the combined population of all the races we put outside of that box. Even though populations of different specific geographical origins do often have characteristic biological traits, the social concept of "race" is not strongly correlated with geographical origin.
That wasn't the whole problem. The literacy tests were only administered to black voters in the first place because of a "grandfather clase"--anyone whose father and grandfather both had the right to vote had the right to vote automatically, and since the literacy tests were only implemented at the end of slavery, in practical terms they only applied to the black population. Naturally, since the test itself wouldn't disenfranchise whites (except for immigrants, but fuck them too) they were free to make it as difficult as possible.
So, that's another example of that cultural bias but it demonstrates we don't need something "complex" to make it stand out.
PS:Sorry for this atrocious use of english, it has been a long day.
That is to say that general knowledge spread btwn cultures (ethnicities) was less pronounced than non-culture based tests (such as abstract concepts). Again, I don't know how valid that conclusion was.
This is an excellent example of the way IQ tests can indeed succumb to cultural training effects. It's pretty clear that this is what's going on with the Flynn Effect, for instance, which is not a real increase in intelligence since not g-loaded:
http://psychology.uwo.ca/faculty/rushtonpdfs/2010%20Editoria...
It's also why Raven's matrices don't work as well as they used to (IMHO). I think my own childhood IQ score is 10-20 points too high, because I'd seen too many IQ-test-like problems.
So, certainly a problem with all IQ tests. But a minor problem, or an existential major one? If it was a major one, these tests would lose their predictive validity - which remains quite high.
What we need to ask ourselves here in the US is why some children get left behind culturally, or why some children live in disadvantaged environments, broken homes or single-parent households. Solving those problems would quickly close the IQ gap by giving more equal opportunity to all children.
http://www.nytimes.com/2009/10/21/us/21latina.html
As I said, the confounding factors exist, but they seem relatively minor. Attempting to effectively homogenize the diverse human gene pool through education is anything but a new effort. If it was possible to educate the Mbuti pygmies of the Congo into the cognitive equivalent of Great Neck Jews, don't you think someone would have tried this already?
Imagine that the remedy proposed was not educational, but rather pharmaceutical. Someone's selling you a drug that purports to turn Pygmy populations (mean IQ 55 or so - but let's be generous and add 10 points for cultural bias) into Ashkenazi populations (mean IQ 115 or so).
You might ask: has this drug ever been tried before? Is there any evidence that it can work? And what happens if the Beastie Boys take it - do they become the world's leading physicists? These would all be very rational questions.
Who would try that? The benign, wise rulers of the Congo?
http://www.time.com/time/printout/0,8816,866343,00.html
In the Belgian Congo last week massed tom-tom drummers practiced a welcome tattoo. Prosperous Negro shopkeepers climbed up wooden ladders and draped the Congolese flag (a golden star on a blue field) from lampposts and triumphal arches set up along Boulevard Albert I, the spanking concrete highway that bisects the capital city of Leopoldville. In far-off mission churches, encircled by the rain forest that stretches through Belgian territory from the Atlantic to the Mountains of the Moon, choirs of Bantu children rehearsed the Te Deum. African regiments drilled, jazz bands blared in...
If you have a Time subscription (yeah right), you can read the whole thing on line. Otherwise, suffice it to say that (a) the memory hole is a lot deeper than you think, and (b) it's physically possible to teach Bantu children (and perhaps even pygmies) the Te Deum. Not clear that it's a safe and effective method for turning them into the King Ad-Rock, however.
The most interesting, and certainly the most effective, attempt to convert a Stone Age population directly to a civilized lifestyle wasn't even in the 20th century. Or the 19th. Consider the Reductions of Paraguay:
Although I suppose the Asian adoptees available, as probably with all races, weren't exactly the pick of their gene pool. Blood runs true, you know - on average.
(Ironically, the perspectives on human heredity generally held in the premodern era are generally more accurate than those taught in modern schools today. The traditional ideas of "blood" were after all consistent with a broad cultural understanding of animal husbandry, and the mechanisms of heredity aren't any different in humans than in pigs.)
I think you illustrate the danger of making assertions without being entirely informed - that is to say that what you state here is the beginning of the truth, but taken out of context. There was (and in many cases still is) a heavy cultural bias built into standardized IQ testing. Traditionally, most tests were developed and tested against middle-income, white children. Black children traditionally tested one standard deviation below their white counterparts, even when controlling for income, etc. However, many experts agree that this is due to cultural bias and environmental factors, agravated by the strong corelation between income and where you live geographically in the US, including resources (or lack thereof) in corresponding school districts. Example: children in troubled, inner-city schools would often have less access to rigourous IQ testing when compared to their urban, middle-class counterparts.
There is also the complicating factor of single-parent or broken homes and the roll that plays in a child's lack of overall intellectual development. This also correlates to income and geographic area (i.e. innercity versus urban environment).
> I've never understood why eliminating one of the few objective measures and making the hiring process entirely subjective was an improvement.
That's another subject and worth talking about. I'm simply stating that you should be very careful when talking about the black-white gap in IQ scores. There are too many factors at play to make a clear case for IQ differences between races.
For an informative article with really enlightening references, see this: http://theafrican.com/Magazine/IQ.htm
"It has been suggested that various aspects of the way tests are formulated and administered may put African Americans at a disadvantage. The language of testing is a standard form of English with which some Blacks may not be familiar; specific vocabulary items are often unfamiliar to Black children; the tests are often given by White examiners rather than by more familiar Black teachers; African Americans may not be motivated to work hard on tests that so clearly reflect White values; the time demands of some tests may be alien to Black culture. (Similar suggestions have been made in connection with the test performance of Hispanic Americans, e.g., Rodriguez, 1992.) Many of these suggestions are plausible, and such mechanisms may play a role in particular cases. Controlled studies have shown, however, that none of them contributes substantially to the Black/White differential under discussion here (Jensen, 1980; Reynolds & Brown, 1984; for a different view see Helms, 1992). Moreover, efforts to devise reliable and valid tests that would minimize disadvantages of this kind have been unsuccessful."
From pp. 93-94 in "Intelligence: Knowns and Unknowns" (1996), Ulric Neisser, et al. Report of a Task Force Established by the American Psychological Association
http://internal.psychology.illinois.edu/~broberts/Neisser%20...
I may not be understanding your argument properly, but doesn't IQ _try_ to measure intellectual development? I understand cultural bias being problem and that makes sense --but this doesn't make sense to me. "The problem with measuring intelligence is it measures intelligence?" What difference does the lack of intellectual development have to do with whether it was genetic, educational, physiological, mental, etc?
I'm not saying banning them completely the right answer here, but putting too much faith into these tests can be dangerous too.
The between the lowest and highest score I ever received on an IQ test is 45 points. That really doesn't sound like an objective measure of anything much (The lowest was when I was stressed, running on no sleep, the test was sprung on me with no warning and had US cultural aspects when I'm from the UK - things like talking about quarters and dimes. The highest was when I had practiced math & logic problems to win a bet that I couldn't add ten points to my IQ :-)
I find the idea of anybody using an IQ test as a factor in employment laughable. As far as I'm concerned any company using an IQ test is hanging a big red sign on the front of their organisation that says "we don't understand what kind of employees we want".
Now, now now. You've forgotten Political Logic.
Political Premise: There are no significant differences in intelligence between blacks and whites. Empirical Premise: Blacks score lower than whites on IQ tests. Conclusion: IQ tests are biased.
The first rule of Political Logic is to never question the political premise!
A tutorial in basic logic is more in order.
The taboo is a completely understandable and largely inevitable reaction against the open and largely fallacious racist claims of academics in the recent past, but it's still a taboo that inhibits honest inquiry. Perhaps in another 50-100 years.
This says nothing about whether an IQ test is biased when used for other purposes than predicting future achievement (such as a job interview), and a supposedly-but-not-actually-objective test is a very dangerous thing; there's a lot of ugly historical baggage where science has met race in the past.
there's a lot of ugly historical baggage where science has met race in the past.
There is indeed:
http://johnhawks.net/weblog/topics/meta/gould-morton-lewis-2...
If that isn't ugly, what is?
http://www.plosbiology.org/article/info:doi/10.1371/journal....
Here's how a slightly more partisan biologist (Peter Frost) responds to the same material:
http://evoandproud.blogspot.com/2011/07/arroseur-arrose.html
Here's the NY Times:
To be honest, your comment reads strongly as though you are projecting.
The English wasn't English English and it frequently used words that sounded like American sitcoms and which, in the context of a test where one answer is write and another wrong, I was quite unsure of the distinctions between.
I guess I scraped through by luck.
I did read the whole thing. She "fought" by answering more questions demonstrating her aptitude, and her interviewers eventually conceded the point:
They tried various other questions, questions from other tests, questions they used for more advanced candidates. Gwen answered as best she could. The men were amazed. To their credit, once they became convinced that she hadn’t faked her results, they knew she would be a great hire. They recommended her for training as a programmer analyst, the most senior position being filled.
Let's pretend there was another black women who wasn't in the 99th percentile, but only the 85th. Let's pretend they would hire anyone over 80th percentile. They might, again, think she was cheating, and hence ask her more, harder, questions, like Gwen here. And since this person is in the 85th percentile, she might not have gotten all the questions right. The examainers would then take this as "proof" that she cheated and she'd be sent home.
We should be worried about the people we don't hear about.
> They tried various other questions, questions from other tests, questions they used for more advanced candidates. Gwen answered as best she could. The men were amazed. To their credit, once they became convinced that she hadn’t faked her results, they knew she would be a great hire.
The story says nothing about whether or not she got all of the followup questions right or not. I think even someone in the 85th percentile should be able to demonstrate they're in the 85th percentile with enough questions--in fact, it might even be easier than for the 99th percentile person because you're assuming that the difference between 85 and 99 is how many hard questions are answered. The first rule of cheating is: never go for 100% accuracy, it makes you stand out.
Companies still look at education, employment history, and ask questions that are trying to evaluate intellectual skills, interests, and abilities. Most of these will correlate with race pretty strongly, too. The entry-level job market is almost exclusively about checking for "good schooling and education".
What's particularly bad about IQ as compared to other measures of employability? Is it just that it's measuring skills that are non-job-specific, or is it actually more racially biased than other evaluation criteria?
I would naively think that IQ tests would be at least somewhat less sensitive to educational status than, for instance, looking directly at someone's educational background. In that case, would you similarly argue that looking at the "schooling" section of a person's resume is a way to claim not to be racist, and then be racist?
Is this a conspiracy theory? It sounds a lot like a conspiracy theory to me. Just sayin'.
Conspiracies do happen. However, when you're trying to persuade me of a conspiracy theory, the burden of proof is high. I expect links - lots of them - and good ones. Try to avoid linking to frauds:
http://blogs.nature.com/news/2011/06/did_stephen_jay_gould_f...
FYI, the most accurate IQ tests available at this time were Raven's matrices:
http://en.wikipedia.org/wiki/Ravens_Progressive_Matrices
I think these have lost some reliability because of a training effect from the general popularity of cognitive brainteasers (probably the cause of the Flynn effect). However, it would be interesting to hear someone explain how they were designed by racists to be racist. (Wikipedia tells me they were originally the Coloured Progressive Matrices, which does sound mighty suspicious...)
At the risk of dragging this thread back onto the topic of the story...
The story describes a number of people who put obstacles in Gwen’s way. None of them are described as being malicious, none of them got together and “conspired.” Some of them genuinely thought they were acting in Gwen’s best interests.
So if someone says that IQ tests were set up in such a way as to create bias, perhaps we can believe that yes they were biased, and possibly deliberately so, without thinking about conspiracies or racism, just people attempting to do what they thought at the time was in everyone’s best interests.
If I say, "Harvard admissions are set up in such a way as to admit large numbers of Jews," someone believing in a Jewish conspiracy will hear my dog-whistle and agree with me. But to someone not believing in a Jewish conspiracy, I can deny everything. A highly useful and malignant propaganda device.
And when you add "possibly deliberately so," you prove this point completely. What else could "deliberately so" mean? So, it may be a Jewish conspiracy that Harvard admits lots of Jews. But it might not be. That's some compromise position.
Normally, when we think of conspiracy theories (like Holocaust denial), we classify Holocaust deniers and "Holocaust agnostics" in the same bin. And rightly so. I'm certainly not interested in compromising on the position that maybe the Holocaust happened, but maybe it didn't.
Also, when I believe things, I prefer to believe them on better evidence than "perhaps we can believe." Sorry - I know it's an inspiring story and you're not looking for a flamewar. This is Asperger-infested HN, however. Geeks like their facts cold, hard, and entirely factual - don't they?
And I choose to believe that you are capable of adding more signal and less noise. On very poor evidence, so far.
It's true that these charges are conventional in our society. Most people would let them pass without notice. However, most people would not endorse letting the unacceptable pass without notice - so I don't think I'm being antisocial.
Sort of. Those of us who live in the real world, where there are shades of grey around every corner, and where subtlety and nuance count for a lot, have realized that we can't always have things served up to us like that.
Unfortunately real life stubbornly refuses to bend to the will of us geeks (and yes, I definitely am one) who want things to behave in ways that can be modeled strictly by cold, hard facts.
Actually, in the past Harvard deliberately restricted the number of jews they admitted: http://www.jewishvirtuallibrary.org/jsource/anti-semitism/ha...
On the contrary, it was probably a mixture of ignorance and 'common knowledge.' I'm going to assume that if you're posting on this site that you probably are involved with making websites or software. Were you aware that a portion of your customers are colorblind in some way? Did you actively design your site to be hard for them to use? Probably not. Does that mean it is a non-issue? No. Much in the same way, white male test-makers make IQ tests that they think "anyone should know," without realizing that capable people with different backgrounds would not score very well at all.
>However, it would be interesting to hear someone explain how they were designed by racists to be racist.
No one in this thread has claimed this. Stop trying to invent a polemic.
It just turns out that whatever IQ does measure correlates heavily with income and where you were born which in turn correlates with race.
Yes, also with high job performance. Actually, IQ correlates with income regardless of family background. That's exactly why it's useful.
I tried to pry the source from JSTOR but failed miserably so here's a crappy pdf copy of "Income Inequality and IQ" http://www.aei.org/book/society-and-culture/citizenship/inco...
Note that this comment was posted before raganwald wrote "and possibly deliberately so," above. So it's not as obtuse as it looks.
As an example of how IQ test results aren't too reliable: IQ tests have been rising year on year for decades (the Flynn effect). Either we're all a pile of geniuses (which makes you wonder what score people in the middle ages got), or IQ test is inflenced by external things.
The measure we usually associate with "innate intelligence", the so-called "G-factor", is actually defined as the correlation between several different factors (education, grades in several subjects, various standardized tests including IQ, career success, socioeconomic status) that all trend in the same direction. It's basically a PCA on the observed metrics. There's no way to measure intelligence directly or even know for sure that it exists as a separate innate "thing" - all we know is that a large number of observable metrics are positively correlated, and then we give a name to the correlation.
What g really means is that you can't design a meaningful intelligence test that doesn't correlate with all the other intelligence tests - with slightly divergent axes for verbal and spatial/mathematical skills. Thus, for instance, backward digit span correlates with Raven's matrices, even though these tasks have nothing obvious in common. Someone who's good at one will be good at the other.
It's also very hard to produce training/educational effects that show an effect on g, though dual N-back perhaps has some promise:
http://en.wikipedia.org/wiki/N-back
However, these kinds of brain exercises have very little to do with education as we know it. The obvious null hypothesis is that we're looking at a physiological effect, such as the quality of myelin insulation in neurons.
Obviously, we see the same correlation effect in CPU benchmarks - any benchmark at all will reveal that a Xeon is faster than a Celeron. The obvious null hypothesis is that the Xeon has smaller transistors and more of them. The causality behind neurological g is probably something just as crude and straightforward.
One could argue, however, that when we compare Xeon motherboards to Celeron motherboards, we see faster DRAM and the like. Perhaps it's the fast CPU's environment, rather than its lithography, that makes it faster.
But... this isn't an argument anyone would make without a strong prior conviction that all CPUs are created equal. It's unclear where such an idea comes from in the case of the human brain, but it doesn't seem evidentiary in nature.
But the causal link between parents' income and kid's academic success is small to non-existent[1] so even if there are factors influencing IQ, wealth doesn't seem to be one, at least not in this way.
[1] "Low Family Income Not a Major Reason For Poor Student Achievement" http://educationnext.org/low-family-income-not-a-major-reaso...
Actually, if you read the whole thing, they didn't believe her "demonstrated aptitude" because of their "prevailing prejedices". She had to fight to get them to see her demonstrated aptitude. Many people would be tired of fighting and would have given up then (or earlier in the process). If the applicant was in the privileged position, they would not have had to fight.
it is illegal in the U.S. to give IQ tests to job applicants
That's because in the past, the IQ tests were not fair, objective and colour-blind, but were set up in such a way as to bias against poorer people (which will be mostly black in the USA). This is easy to do with IQ tests, just ask lots of questions that require good schooling and education (e.g. word questions).
Companies used to use IQ tests, and hence claim not be racist, and then be racist. The ban on IQ tests in job interviews stems from this US Supreme Court case http://en.wikipedia.org/wiki/Griggs_v._Duke_Power_Company and has hence probably improved equality.