Hacker News Reader

Comment by User23 | Hacker News Reader

The whole IQ is only one measure argument really misses the main lesson from psychometrics. It’s entirely well known that IQ test taking is a specific skill. However it’s also known that there is a g factor that predicts about 50% of performance on basically all tasks with a cognitive component. The other 50% is indeed skill specific. So someone with a slightly above average g factor could still be a world class composer or quarterback. Contrariwise someone with an extremely high g might be very good at many different tasks, but not world class at any. So when one talks about measuring intelligence in a general sense, one ought to be talking about g.

View on HN

The entire field of Psychometrics is suspect and is a pseudoscience. It came into existence in the early days when ideas from Social Darwinism, Racism and Eugenics were mainstream in the Western world. People ascribed an importance factor to traits which they considered successful in their culture/society/environment and then started making universal and general inferences from it without regard to its validity. These were then codified as tests, published and propagated to the detriment of all.

Today we know better; Humans are highly adaptable and everything only exists in the context of a specific environmental/cultural/social context. Genetics vs. Epigenetics, Nature vs. Nurture, Nutrition, Economic strata etc. are all inputs when it comes to our "fitness" for a particular task. We play "roles" which are context-dependent from which you cannot generalize to the "whole person".

Psychometrics has been called a "Pathological Science" - https://en.wikipedia.org/wiki/Pathological_science

This article named The Myth of Measurement brings a bunch of research together to show why - https://paulenglert.com/the-myth-of-measurement/

If the field makes testable predictions that prove true then it cannot be a pseudo science.

And indeed, one can predict with accuracy better than chance how well someone will perform in a software developer job (for example) from an IQ test.

This same test may do less well in predicting the ability to hunt an antelope across the serengeti, but that still wouldn't make psychometrics pseudoscience.

Also, if there wasn't a g factor that would be fairly weird, as there are numerous little slightly deleterious mutations one could have that can minorly affect neuronal functioning. You'd genuinely expect there to be a decent correlation here.

When the characteristics that a test is supposed to measure are completely ill-defined (there is no concept of a "Measurable Intelligence" independent of a context, leaving aside biological retardation) they are a measure of nothing definitive and any correlations/predictions are subject to bias and/or randomness.

Here is Nassim Taleb;

1) IQ is largely a pseudoscientific swindle - https://medium.com/incerto/iq-is-largely-a-pseudoscientific-...

2) Fooled by Correlation: Common Misinterpretations in Social "Science" - https://www.academia.edu/39797871/Common_Misapplications_and...

3) Correlation videos - https://nassimtaleb.org/tag/correlation/

See also :

a) The Pseudoscience of Psychometry and The Bell Curve - https://www.jstor.org/stable/2967209

b) Pseudoscience and Mental Ability: The Origins and Fallacies of the IQ Controversy. - https://eric.ed.gov/?id=ED164662

Out of curiosity are you espousing that differences in intelligence don't actually exist, or just that we need to come up with better ways to measure them?

I wonder how many Fields medalists score one or two standard deviations below average on the Stanford-Binet. If the measure is meaningless then roughly half should, right?

Also while I certainly respect Mr. Taleb's insights and the mass market appeal of his punchy authorial persona, based on various writings of his I conclude that he views intelligence very narrowly in the "good at trading options" sense, which coincidentally is an area where he has demonstrated ability. Funny that. And to be fair in my own trading I benefitted quite a bit from his maxim that the first goal is survival.

I'd bet that Fat Tony would probably score well above average on an IQ test too, he just wouldn't be caught dead taking one. And I don't blame him; the kind of people who mistake aptitude for achievement are generally insufferable.

> Out of curiosity are you espousing that differences in intelligence don't actually exist, or just that we need to come up with better ways to measure them?

Yes and No to both.

Simply put;

1) There is nothing called "General Intelligence" beyond what enables a Human to survive in his/her environmental niche (including cultural/social since we are social animals). Leaving aside obvious biological retardation, if Extraversion/Introversion, Agreeableness/Disagreeableness, etc. enables one to survive, they are equally "Intelligent". Trying to rank these attributes on a ordinal scale and do quantitative calculations on them is nonsense. See Stephen Jay Gould's The Mismeasure of Man (https://en.wikipedia.org/wiki/The_Mismeasure_of_Man) for convincing arguments.

2) Humans are highly adaptable and exhibit dynamic behaviour in different contexts. This "role" and hence the "exhibited intelligence" is specific to that context but if that adaptation subsumes the need for another related context then obviously one will be intelligent in both the contexts. For example London Taxi drivers actually grew their hippocampus and became more intelligent in both spatial navigation and memory recall - https://www.scientificamerican.com/article/london-taxi-memor...

3) We know that gene expression can be influenced by the environment via Epigenetics over generations (see the documentary The Ghost in your Genes). Environment (cultural/social/nutritional/economic) now becomes of paramount importance and thus a person's "fitness" for a role is dependent on Nature, Nurture and his unique Family Past history. He will now be intelligent in certain roles not so in others.

4) Wrong understanding and misuse of Statistical methods in the above scenarios. This is where Taleb's writing/videos help one to separate the wheat from the chaff.

- Notes on Taleb's article - https://supermemo.guru/wiki/Taleb:_IQ_is_a_pseudoscientific_...

- IQ is a dismal measure of intelligence - https://supermemo.guru/wiki/IQ_is_a_dismal_measure_of_intell...

> See Stephen Jay Gould's The Mismeasure of Man

No thanks. Why would I refer to a source whose author either lied to get the results that suited his politics or was completely incompetent? From the NY Times[1]:

  But now physical anthropologists at the University of Pennsylvania, which owns Morton’s collection, have remeasured the skulls, and in an article that does little to burnish Dr. Gould’s reputation as a scholar, they conclude that almost every detail of his analysis is wrong.

Frankly I have my doubts about your knowledge of the subject if you’re relying on thoroughly discredited sources.

You did nicely restate Mr. Taleb’s novel redefinition of “intelligence” as survival ability though. No doubt then Taleb believes the humble tardigrade is one of the most intelligent creatures on Earth.

[1] https://archive.is/GhiVK

I pointed you to the wikipedia page of the book for a reason. The fact that you failed to note that the study you quote has itself been debunked/criticized by later studies leads me to question your knowledge of the whole topic.

Read the section on Reassessing Morton's skull measurements and follow the links as needed.

Finally, you might also want to checkout the following;

1) Triarchic theory of Intelligence - https://en.wikipedia.org/wiki/Triarchic_theory_of_intelligen...

2) Theory of multiple Intelligences - https://en.wikipedia.org/wiki/Theory_of_multiple_intelligenc...

The critique amounts to essentially: well Gould is observably incompetent or a liar, but Morton was a racist so Gould is somehow still right.

As for the triarchic theory, it’s not even wrong. I’m completely uninterested a non-empirical “theory” of intelligence.

You are in full-blown denial mode of every evidence/theory contrary to your cherished "g factor"/single general intelligence/psychometrics "belief system".

I will end this with the following observation;

In this modern day and age anybody who believes;

a Bedouin from the Sahara, a Aborigine from the Australian Outback, a Chukchi from the Northern Arctic, a tribal hunter from the Amazon Rainforest, a nomadic herder from the Mongolian Steppe, a Sherpa from the Himalayas, a Xhosa from Africa and a Techie from Silicon Valley

can all be ranked on a single scale using a single metric of "General Intelligence" is ... manifestly wrong and unscientific! The absurdity is self-evident.

Now you’re just falling back on bluster. You’re not even making claims anymore.

You may as well argue that height isn’t a scientific metric because some people have longer legs and others have longer torsos so clearly any single height metric is meaningless. It’s pure flimflammery.

In any event it’s pretty clear that your beliefs on the subject of intelligence are not evidence based, so I don’t think further discussion will be intellectually satiating for either of us.

You don't need an exact definition in order to be able to measure something approximately.

In fact, one could taboo the word "intelligence" entirely and merely say "that thing we observe where people who do better at these tests seem to also do better at these jobs and are more likely to win board games[1].". While the tails do come apart eventually, the correlation is observable up to at least 130 (there was a strong correlation at my previous work place between those who did well at board games and those who got a lot of good technical work done, these people also tended to have +2sd test scores compared to the +1sd average).

Btw: I've read Taleb's charged essay before and found it rather unconvincing. Very much "given these particular papers it's possible high IQ scores correlate with nothing other than high test scores".

[1] I'm not just basing this off my own observations, good chess players seem to have 130 level IQs: https://www.sciencedirect.com/science/article/abs/pii/S01602...

https://news.ycombinator.com/item?id=36862687

Odd then that intelligence psychometrics is far less affected by the replication crisis than the other social sciences. The person that invents a useful intelligence test that doesn't display any politically inconvenient bias is going to make a whole lot of money.

On the importance or supposed profundity of the g factor, here's Cosma Shalizi:

http://bactra.org/weblog/523.html

(If you like, skip to "Correlations explain g, not the other way around").

Interesting read. I’m not inclined to critique it much beyond observing that the author clearly set out with an objective in mind and argued well to that end. My brief response is that the fact that we can find similar patterns in noise doesn’t actually show that the discovery of patterns indicates the data is noise.

And of course intelligence is a real and observable thing. Some people really are considerably smarter than others. They have better recall, learn faster, and are able to reach correct conclusions either more quickly or with less information than less intelligent persons. Reducing that difference to a single factor almost certainly is a simplification, but just as we can usefully talk about height without getting caught up on difference in leg, torso, and neck length, we can talk about intelligence similarly.

The article addresses this: for instance, if you were to create a "sound mind and body" score that included "hot or not" results, and the results of the Presidential fitness test, you would inevitably come up with a score that was even more predictive of life outcomes than IQ or g.

Physical attractiveness correlates with many good things like fertility and genetic health. Not to mention beautiful people benefit from the halo effect. The causality there though is likely that the relative superiority of attractive persons causes the halo effect to exist. There’s a pretty obvious set of mechanisms that offer causal reasons why attractiveness leads to better outcomes.

As for the presidential fitness exam, well that’s true but quite facile. Much like how people who have taken the Wonderlic test are considerably more likely than the general populace to be or have been in the NFL.

In fact, I wouldn’t be surprised if taking aptitude tests at all regardless of score predicts above average outcomes because of a selection effect.

The point is that you can drag a lot of unrelated things together and get an increasingly predictive metric, despite the underlying factors being unrelated, which simply means that "being predictive" is not by itself dispositive of some underlying mechanism. It doesn't mean prediction isn't real! The problem is people going from "prediction" to "inferred ground truth" --- which is exactly what happens in the "g" discussions.

> which simply means that "being predictive" is not by itself dispositive of some underlying mechanism.

Right. I don’t see how any statistical analysis can tell you if a mechanism exists or what it is. Quite the opposite. The better you understand the mechanism of the process that generates your data points the higher quality your statistical analysis will be. And in the case of intelligence we know there must be an actual real underlying factor because some people are observably smarter than others.

I don't think anybody is arguing that people aren't observably smarter than other people; the question is about the metrics we're using, which, I mean, an equally vivid (and equally non-rigorous) argument would be: if you really buy into the IQ/g story, go hang out with a bunch of Mensa people and see how long you hold onto that belief.

I’m willing to bet that a Y combinator founder’s event’s attendees have as high if not higher average IQs than persons at a Mensa meeting.

Mensa appears to select for high aptitude and low achievement, which is a really unfortunate combination.

Reply on news.ycombinator.com