Against Full Scale IQ Fetishism

Why it is entirely possible Scott Alexander and Kasparov's IQ is 135

There was a recent controversy over me having estimated the IQ of Scott Alexander at 133. Many found it comical that such an accomplished and talented writer could have an IQ so “low” - when in reality it is quite possible this is his real score.

Keep in mind that intelligence and FSIQ are not equivalent - intelligence refers to an agents latent general cognitive ability, while the FSIQ is the standardized sum of the scores on cognitive subtests. These variables correlate at about .92 (n = 485) (Farmer et al., 2014), so most people’s FSIQ scores correspond closely to their latent general intelligence, but it is possible to see discrepancies.

I don’t intend to be another idiot who tries to challenge the solid consensus of intelligence research: that IQ tests are highly reliable, rarely culturally biased, highly predictive of academic outcomes, and moderately predictive of life outcomes. Instead, I would like to contest the position that many hold about intelligence: that it is extremely predictive of life outcomes and the accomplishment of elites and eminent people must be highly dependent on their IQ. This is particularly apparent when people evaluate intellectuals; they uncritically assume that the quality of their work must be due to their extreme levels of intelligence.

As many of you know, intelligence and success tend to go together. The following table from Strenze’s 2015 meta-analysis concisely shows this. I must note that the correlation between income and intelligence is about .36 (n=8499) in the NLSY79, and .3 (n=4065) in the NLSY97, so I’m not entirely sure why the correlation is so low in the meta-analysis. Regardless, intelligence appears to predict many different life outcomes, though the correlations are low enough for there to be room for other sources of influence such as luck, personality, or specific cognitive abilities.

It is almost certain the effect of intelligence on success is causal - analysis that control for SES or use sibling controls show the same associations (Murray, 1994; Murray, 2002; Hegelund et al., 2018; Hegelund et al., 2019).

Intelligence correlates at .39 with work sample performance after correcting for the unreliability of both work samples and supervisor ratings. Currently a prominent critique of this literature is circulating, but I find that it falls flat. For instance, the authors of the critique implicitly advocate for correcting for intra-rater reliability instead of inter-rater reliability:

Hunter and Hunter (1984) assumed a reliability of 0.6 for their corrections, which some investigators have considered to be too low (Hartigan & Wigdor, 1989). Bertua et al. (2005) used the same figure for their meta-analysis of British studies. Moreover, that estimate was based on inter-rater reliability. Murphy and DeShon (2000) pointed out that differences between raters should not be considered error to be corrected because different raters may be looking for different things in a worker. Instead, intra-rater reliabilities should be used. However these tend to be much higher: 0.86 rather than 0.6. according to the meta-analysis carried out by Viswesvaran, Ones, and Schmidt (1996). The lower the value adopted, of course, the bigger the inflation to raw correlations. Using the reliability of 0.6, for example, inflates the correlations by 29%. By comparison, distinguished statistician John Hartigan, and colleague Alexandra Wigdor, favour the 0.8 estimate which only inflates the correlation by 12% (Hartigan & Wigdor, 1989) As Murphy (2003) says, evidence of error is so pervasive that many commentators urge caution in using supervisor ratings as criterion of job performance.

That is, that the correlation between job performance and intelligence should be adjusted based on the reliability of a single rater, instead of the reliability between different raters. This is ridiculous; who knows why different raters rate workers variably in job performance? For all we know, John could be rating Linda a 10 because she sucked his dick last night, while Ed might be rating her a 4 because she reported him to HR for groping her. These are not meaningful differences in job performance - as they were given for reasons unrelated to it.

Estimating IQ based on job performance

If you have the correlation between occupational success and intelligence and an individuals occupational success within a job, you can come up with a crude estimation of intelligence. In fact, many of the individuals who were criticizing my estimate of Scott Alexander’s IQ were pointing out his success in writing as a refutation of my estimation of his IQ at 133. However, this argument only works if you think being a good quality intellectual is highly g-loaded (r > .5) and that he is a great (z-score > 4) intellectual.

If you don’t think being a good intellectual is highly g-loaded (0.2 < r < 0.4), it doesn’t matter how good of an intellectual Scott Alexander is - you won’t predict his IQ to be above 142. On the other hand, if you think being an intellectual is highly g-loaded, then your estimate for Scott’s IQ could be as high as 166.

I do think Scott Alexander is a great intellectual - he obviously knows what he writes about, is a talented writer, is willing to talk about controversial issues, and has been consistently pumping out good posts for a long time. However, I don’t think what he is accomplishing is highly dependent on intelligence.

How much does IQ contribute to quality as an intellectual?

While it is a little late in the article to do this, I think that a definition for a intellectual quality is in order. I would define it as a composition of the following variables:

• Ability to distinguish between truths and falsehoods.

• Rigor of work (e.g. citing and reading sources, willingness to dive in the weeds).

• Quantity of work.

• (Less important) Aesthetics (e.g. presentation of work, writing style).

From a non-scientific standpoint, intelligence to some degree should contribute to intellectual quality. If you consistently make reading comprehension errors and do not understand mathematics, you are going to have a hard time ascertaining what is and is not true in complex topics. Indeed, Joseph Bronski (2022) [video] has estimated that roughly 75-95% of humans cannot have political agency because of deficiencies in reading comprehension and mathematical understanding. By political agency, he means the ability for a citizen to recognize political problems and advocate for adequate solutions to them, which is roughly what an intellectual is doing in the political sphere.

Beyond this threshold of about 115IQ, I think intelligence would still be predictive of quality as an intellectual. This is because higher levels of general intelligence would enable you to understand concepts faster, reducing the time constraint of creating good output.

Given the IQ scores of Nazi top brass and top scientists, the idea of a necessary threshold of intelligence for eminence seems plausible:

From the Making of a Scientist:

Verbal ability of eminent scientists:

The median score of this group on this verbal test is approximately equivalent to an IQ of 166. The lowest and highest scores would be at about the levels of 121 1Q and 177 IQ.

Spatial ability of eminent scientists:

These scores can also be translated, as were the scores on the verbal test, into approximate IQ terms. The lowest score, in these terms, is about the same as the lowest score on the verbal test; it gives an IQ equivalent of 123. The highest score on this test, however, does not quite reach the median score on the other, being equivalent to an IQ of 164, while the median IQ on this test is 137.

Mathematical ability of eminent scientists:

Let us look at the equivalents on this test. It is not correlated with age (the correlation coefficient is .00). The lowest score on this test is about equivalent to an IQ of 128, the median score to an IQ of 154 and the highest to an IQ of 194. That is very high indeed. Mathematical ability is certainly important for work in physics, but it seems that it can also be important in some other sciences, particularly biology and psychology.

Nazi top brass IQ scores:

I find very few scores below 115 across the board, supporting the hypothesis that high levels of intelligence are necessary for being a political, cultural, or scientific elite.

What qualities would matter besides intelligence?

To properly accomplish intellectual quality, an individual must have the following characteristics:

• A reasonably elevated amount of intelligence

• Willingness to invest time in research and writing to create good output.

• Prioritizing discovering truth over money, political alliances, and fame.

• Willingness to tackle controversial or inflammatory issues despite potential social pushback.

• (Less important) Connection to high quality communities (HBD, rationalist).

• (Less important) Absence of connections to low quality communities (right wing grifters, wokes, communists).

Only one of these qualities is highly g-loaded - the rest are either personality traits or are factors of luck. Willingness to invest time is a measure of drive, prioritizing truth is a matter of virtue, willingness to tackle controversial issues is a facet of disagreeableness, and connection to communities is a function of self-selection and luck. That is not to say these personality characteristics are not even associated with intelligence, but that they are largely independent from it.

I think the less important qualities can help make an individual overcome a deficiency in one of the 4 main areas, but they cannot make a gold coin from a piece of shit. To my knowledge, the rationalist community has failed to reject Nick Bostrom for his infamous email - I think this sets a precedent that enables people who would otherwise be afraid of engaging with radioactive topics to do so. This is much more difficult to do if you are surrounded by low quality people who will seethe at any dissenting voice.

I personally think the time and drive constraints are the biggest filters to intellectual quality. Very few people have the time necessary to spend researching a wide range of topics or the drive to do so, cutting off the amount of people who can be a “great intellectual” in a very obvious and intuitive way.

In terms of a case study where an individual fails to achieve greatness due to a deficiency in one of these four qualities, I present to you the following exchange between The Distributist and Kevin Bird, where Kevin Bird admits to not caring about truth.

(See extended version in Appendix for wider context)

The Distributist: Why do you believe it (it = racial equality and harmony) is possible?

Kevin Bird: I don’t view the historical occurrences as evidence that it’s not possible I just view them as evidence that it hasn’t happened.

The Distributist: You believe it’s possible to do this because the past isn’t an indication abvout how the future behaves.

Kevin Bird: Yes, exactly.

The Distributist: That’s insane. If we didn’t the past wasn’t an indication of how the future operated we couldn’t do science. You’re violating Hume’s inductive hypothesis. It’s like saying the sun rose every single day, you know, but it won’t rise in the future because we can’t induct on observations. How do you do science with this assumption?

Kevin Bird: I do science because I find it intellectually engaging. To be completely honest…

The Distributist: I said HOW would you do science!

Kevin Bird: I just do it with not as much interest in attaining or discovering truth.

The Distributist: So you’re not interested in discovering truth?

Kevin Bird: No, no. Not really.

Here Kevin Bird shows his lack of a necessary quality: caring about the truth. Whether he has the cognitive abilities or connections to be able to become a good intellectual is irrelevant, because he cannot reliably distinguish between truths and falsehoods in certain domains if he refuses to do so.

What I’m trying to get at is that quality as an intellectual is highly dependent on having several rare characteristics besides intelligence, which implies two things:

1. Very few people are capable of being intellectual elites or having even basic political agency.

2. Being a good intellectual is not highly g-loaded, as various other qualities need to be combined with it.

Emil also highlights that Arthur Jensen has made a similar argument in the past: that genius and giftedness are two different things in his work ‘Giftedness and Genius’. He mathematically models it with this formula:

Genius = High Ability X High Productivity X High Creativity

And before him, Galton advanced the ‘triple event’ hypothesis, that eminence is dependent on three abilities:

The particular meaning in which I employ the word ability, does not restrict my argument from a wider application; for, if I succeed in showing—as I undoubtedly shall do—that the concrete triple event, of ability combined with zeal and with capacity for hard labour, is inherited, much more will there be justification for believing that any one of its three elements, whether it be ability, or zeal, or capacity for labour, is similarly a gift of inheritance.

Empirical evidence suggesting I am correct:

If you thought being a quality intellectual was strongly related to intelligence, then you would expect the most intelligent people to believe the truest things (we don’t do post-modernism here). In most studies done on this topic, the relationship between cognitive ability and political beliefs isn’t strong, and rarely exceeds r = .3.

Noah Carl’s 2014 data:

Meta-analysis of right wing economic beliefs and IQ:

Kanazawa, Why Liberals and Atheists Are More Intelligent:

Some people may bring up the “midwit effect” that is, linear relationships between intelligence and political views may obscure the trends in beliefs that occur in individuals with IQs above 120. I have not seen actual evidence of this happening in any political belief, and I invite those who have it to present it.

Even in cases where there is a definitively true political belief, it can still be negatively correlated with intelligence. I found that in individuals who agreed that the average difference between Blacks and Whites have worse jobs, income, and housing than Whites due to having less-inborn ability to learn were less intelligent than those who disagreed.

Here is the difference in wordsum score (d = -0.83):

Of course I am aware that thinking Black people have worse outcomes in jobs, wages, and housing due to differences in intelligence isn’t necessarily a hereditarian belief, but if you agree with that statement you might as well be half-way down the rabbit hole.

Regardless, statistical evidence seems to suggest that a large portion of differences in income between Whites and Blacks can largely be reduced to differences in intelligence between them (Meng Hu, 2022; Jensen & Nyborg, 2001).

Income by race and general intelligence:

If a political belief is true but is negatively correlated with intelligence, I find it hard to believe that intelligence strongly contributes to believing in true political beliefs. Given that being able to distinguish between truth and falsehoods is a necessary quality to be a good intellectual, this suggests to me that this job is not highly g-loaded.

Appendix

Kevin Bird vs Distributist context:

The Distributist: You’re asking me to believe that something (racial equality and harmony) that happened a lot in the past will not happen in the future because something that’s never happened in the past will happen now. That’s what you’re asking me to believe. Would you bet on that happening?

Kevin Bird: If you have good reason to believe that what you are changing is a driving causal force - what is standing between actually producing some level of equality between majority and minority groups.

The Distributist: Well that’s an interesting question and we will get to that, but that’s not the question I just asked. You’re asking me to essentially bet the future of my posterity which you don’t care about but I do on the fact that something that has historically happened won’t happen in the future but that’s never happened in the past will happen now, even though… Is it happening now? Are we ameliorating the differences between the minority groups that are entering this country? I don’t see that.

Kevin Bird: I don’t see it either…

The Distributist: So you’re not even doing what could possibly ameliorate this right now?

Kevin Bird: Well no, the system that we’re participating in is sort of designed not to do it. It’s the way the system works.

The Distributist: And it’s the way every system works all the way back in human history.

Kevin Bird: I mean… so if we want to talk about….

The Distributist: Has any human society done this in the past?

Kevin Bird: I don’t know enough history of groups going back to even more egalitarian periods…

The Distributist: Isn’t that an important question? Because if it’s never been done in human history, which to my knowledge it hasn’t, then why would you bank a lot on it happening in the future?

Kevin Bird: Because I think it would solve the issues… (talking over each other)

The Distributist: is it possible? If it’s (racial equality and harmoney) never happened in the past, and it’s not happening now, wouldn’t you ask yourself if it is possible?

Kevin Bird: Not entirely. I think for the most part we’ve been acting under a very similar framework.

The Distributist: This is an incredibly incurious position for a scientist to take. You’re saying that a phenomenon that has never happened and you’re not wondering if it’s not impossible? That’s the next logical question any scientist would ask themselves.

Kevin Bird: That’s not my thought process behind it. My thinking is that this is a major cause of the issue, so it’s something that should be attempted to be addressed. And I think the extent to which it could solve the problem…

The Distributist: That wasn’t even the answer to the question. I said whether it’s possible to address it.

Kevin Bird: Yes, I think it is.

The Distributist: Why do you believe it (racial equality and harmony) is possible?

Kevin Bird: I don’t view the historical occurrences as evidence that it’s not possible I just view them as evidence that it hasn’t happened.

The Distributist: You believe it’s possible to do this because the past isn’t an indication abvout how the future behaves.

Kevin Bird: Yes, exactly.

The Distributist: That’s insane. If we didn’t the past wasn’t an indication of how the future operated we couldn’t do science. You’re violating Hume’s inductive hypothesis. It’s like saying the sun rose every single day, you know, but it won’t rise in the future because we can’t induct on observations. How do you do science with this assumption.

Kevin Bird: I do science because I find it intellectually engaging. To be completely honest…

The Distributist: I said how would you do science!

Kevin Bird: I just do it with not as much interest in attaining or discovering truth.

The Distributist: So you’re not interested in discovering truth?

Kevin Bird: No, no. Not really.