Name: Bias in mental testing
ISBN: 9780029164303

What Hacker News says

gwernOct 2015

> For children the test takes, "48–65 minutes to administer" and requires more than pencil-and-paper [1].

For the full-scale WAIS, perhaps. But you can still extract a good estimate from the subscales, and needless to say, the WAIS is far from the only IQ test out there.

> I think doing a regular comparison of the standard IQ test results with this new scanning method could be a good way to account for test bias [2].

Huh? The brain scans are being validated using the existing IQ tests in the first place, and if one doesn't accept all the existing literature on the unbiasedness of current IQ tests (where the concepts are not too difficult - for example, I think many laymen could get through and understand most of _Bias in Mental Testing_, Jensen 1980 http://emilkirkegaard.dk/en/wp-content/uploads/Bias-in-Menta... explaining what biases there would be, how one would test for them, and why IQ tests are unbiased), one is not going to accept the results of the infinitely more complicated & subtler brain imaging pipeline where there's so many details & variability that they had a pretty serious replicability problem for a while, especially when they just yield the same results that the original IQ tests would have.

thread

xyzzyzDec 2021

You speak as if these issues have not been studied to death a generation ago already. Jensen’s “Bias in mental testing”, a nearly 800 pages volume full of references, covers it rather extensively, and it came out 40 years ago.

What makes me self assured here is that best counter arguments to the body of scientific research I read on this topic that I hear is people musing about some hypothetical IQ tests that might possibly ask about baseball. I mean, consider Raven’s progressive matrices. It’s not the best IQ test (for reasons other than knowledge bias), but it correlated very well with WAIS and Stanford-Binet, two most popular and arguably the best IQ test. It asks for no knowledge of baseball, regattas, or anything else, in fact, it used no words at all. How comes that it nevertheless gives the same results as other, allegedly biased IQ tests? What exactly makes it biased, and why in exactly the same direction as other tests?

thread

xyzzyzMay 2019

Word tests are used as part of good, high quality IQ tests (e.g. WAIS and Stanford-Binet, though obviously not Raven’s). They are in fact one of the most useful and predictive parts (I.e. they are most g-loaded). They are biased (i.e. not measurement invariant) if you compare subjects from different countries/cultures, but they are not biased if you compare subjects from the same country/culture. For example, the score on word analogies sections has exact same predictive validity for white Americans as it does for black Americans. See Arthur Jensen’s “Bias in mental testing” for a very comprehensive treatise on the issue.

thread

opticksversiJan 2009

Something I have not seen pointed out yet (in articles and discussions over the preceding many years) is the fact that CAPTCHA tests are essentially Gestalt Completion tests. A Gestalt Completion test is a type of IQ test. As CAPTCHA tests get progressively more difficult to solve, they might weed out progressively larger portions of the population - the human population, that is.

Perhaps, as a side-effect, the quality of web-based discussion will rise.

From Arthur Jensen's 1980 book Bias in Mental Testing (pp157-158):

Gestalt Completion. Familiar words and objects are shown in various degrees of "mutilation" and the subject must infer what the word or object is. Performance is scored according to the average time taken per item or the number of items completed within a given time limit. The difficulty of any given item is a function of its familiarity and the extent of the "mutilation"; these can be varied independently in making up items. ... Gestalt completion items, which are known to be a good measure of the general intelligence factor (as well as of a small group factor labeled speed of closure), merit further experimental investigation in their own right.

thread

DabAsteroidSep 2008

over and over until something works

Indeed. ...Until something accidentally works.

Lower animals tend more to learn by accident. They might memorize or ritualize what works, but they do not resonate with why it works. Higher animals tend more to learn by developing mental-resonance with natural processes (i.e. they do resonate with why things work), saving the effort involved in accidental-discovery learning techniques such as trial-and-error.

The main indices of intelligence in animals are the speed of learning and the complexity of what can be learned, the integration of sensory information to achieve a goal, flexibility of behavior in the face of obstacles, insightful rather than trial-and-error problem-solving behavior, transfer of learning from one problem situation to somewhat different situations, and capacity to acquire abstract or relational concepts.

(Arthur Jensen. Bias in Mental Testing. http://www.amazon.com/Bias-Mental-Testing-Arthur-Jensen/dp/0... From the section "Animal Intelligence".)

Perhaps trial-and-error should be classed as a subset of accidental discovery -- e.g. there would be myriad ways of accidentally discovering, one of which would be the motivation-driven trial-and-error.

Place a chicken on one side of a limited-length chain-link fence, and food on the other side, and the chicken will use the sheer force of its hunger-drive to discover how to get to the food. It has a motivation to get to the food, but it cannot mentally-resonate with the spatial processes -- the distance, the fence, gravity, etc. -- that are keeping it and the food separate. It therefore applies its motivation to accidental discovery -- sepcifically, the motive-driven accidental-discovery technique of trial-and-error. E.g., it will scramble around in a panic until it accidentally discovers one of the two routes to the food -- a few feet to either side of its original location. This process might take many minutes or hours. The chicken might even die of hunger without discovering that walking a few feet around the fence will get it to the food.

In the same test, a dog will mentally-resonate with the spatial processes involved, and simply walk around the fence.

The chicken applies the accidental-discovery technique of trial-and-error. The dog theorizes, i.e. mentally-resonates, i.e. applies science.

thread