In the past week or two, there have been several news stories and blog articles written in regard to a paper that came out this month titled, “Why Liberals and Atheists Are More Intelligent”.
It would be easy for the untrained eye to read such a headline and think, “Gee…. this was published in a peer reviewed journal, so it must really be something…” However, if you read through the paper in detail, and look at the longitudinal data methods used for the analysis, one begins to question the validity of such broad assumptions.
I have run across other studies in the past which claim similar correlations between intelligence and atheism and political affiliation, so I was not at all shocked when I first read the title. What did shock me, however, was the nature of the arguments presented by Satoshi Kanazawa in defense of his position. The problem with papers like this is, when the analysis of the data is so far from logical, it ends up threatening the validity of the entire data set used, even though the data itself may have been relevant and sound.
The information he used for the basis of his analysis came from two sources. They are publicly available data sets, one from the National Longitudinal Study of Adolescent Health (Add Health), and the other from the General Social Surveys (GSS). While I have a few issues with their testing measures (mainly to determine intelligence), those problems are nothing compared to the gross misinterpretations of the data and the absurd theories presented by Kanazawa.
First of all, the measure used to determine intelligence in the Add Health study is a vocabulary test. Yes, I know vocabulary tests correlate with general intelligence, but it is a correlation, not a true measure of general intelligence. It is a measure of crystallized intelligence, or a sum of the learned knowledge at any given point in time.
A true measure of IQ should not be based on how many math courses you have taken, or how many works of literature you have read, or how many vocabulary words you have memorized prior to the examination. The GSS used a Picture-Word Vocabulary Test as their measure of IQ, which is just as bad.
Anyone who has done any amount of research in the field of intelligence knows that the correlations of individual skill tests to overall ability only hold up for a segment of the population, mainly the lower to middle portion of the distribution. They begin to break apart after IQs of 120-125, which means there are other factors involved in IQ that are not measured by those tests.
This means that a majority of you reading this article right now have IQs that are past the point of solid correlation, so these assumptions and correlations are no longer as valid, just by the nature of being out of the group of highly correlated subjects. I have written oodles on this topic before, so I will not elaborate further; if you want a more detailed explanation of the validity of using individual skill test correlations (such as a verbal test) when determining IQ, you can go here.
Now to discuss Kanazawa’s “theories”...
One of the problems I have with many Evolutionary Psychologists is that they fail to understand that we are humans, not animals. We have other things going on cognitively and neurologically as a species that do not apply to animals. First, we have Kanazawa’s own Savanna Principle, which he uses as the foundation for his Savanna-IQ Interaction Hypothesis (sounds impressive, doesn’t it?) which he continuously references throughout the paper as a supposed valid explanation for his conclusions:
The Savanna Principle: The human brain has difficulty comprehending and dealing with entities and situations that did not exist in the ancestral environment.
Just what is considered the “ancestral environment”? The ocean where we were amoebas? A primitive cave? The home of the earliest hominids? The part of the world where we see the emergence of the genus Homo? Aren’t we still discovering more details about the origin of our ancestors? How do we know what that environment was like specifically, and thus draw these types of conclusions?
He goes on to say:
The Savanna Principle can potentially explain why some otherwise elegant scientific theories of human behavior, such as the subjective expected utility maximization theory or game theory, often fail empirically, because they posit entities and situations that did not exist in the ancestral environment. For example, many players of one-shot Prisoner’s Dilemma games may make the theoretically irrational choice to cooperate with their partner, possibly because the human brain has difficulty comprehending completely anonymous social exchange and absolutely no possibility of knowing future interactions (which makes the game truly one-shot). Neither of those situations existed in the ancestral environment, but they are crucial for the game-theoretical prediction of universal defection.
That sounds like a beautifully complex, yet completely unnecessary and unfounded explanation for the result of the Prisoner’s Dilemma. I tend to follow Occam’s razor, which states that the simplest explanation or strategy (that follows and supports the data) tends to be the best one.
Here are my thoughts that might explain why the Prisoner’s Dilemma prediction and game theory might fail empirically:
We are humans, not bots. When it comes to psychological behaviors, we don’t follow algorithms.
We have free choice to make any decision we want, even if it is irrational. In fact, humans are irrational a lot of the time.
We are creatures of emotion. Animals are not. Emotions effect our decision-making in ways that don’t apply to animals. We have feelings of loyalty, trust, and general misanthropy, which all influence our exhibited behaviors. To say that we should act a certain way because an algorithm says it is the most rational choice is completely disregarding the factors that make us human.
After Kanazawa defines the Savanna Principle, he goes on to describe what he calls the Savanna-IQ Interaction Hypothesis:
The logical conjunction of the Savanna Principle and the theory of evolution of general intelligence suggests a qualification of the Savanna Principle. If general intelligence evolved to deal with evolutionarily novel problems, then the human brain’s difficulty in comprehending and dealing with entities and situations that did not exist in the ancestral environment (proposed in the Savanna Principle) should interact with general intelligence, such that the Savanna Principle holds stronger among less intelligent individuals than among more intelligent individuals. More intelligent individuals should be better able to comprehend and deal with evolutionarily novel (but not evolutionarily familiar) entities and situations than less intelligent individuals.
Kanazawa’s definition of intelligence—which he admits strays far from currently accepted theories of intelligence by those in the field—is defined as such:
[General Intelligence] may have originally evolved as a domain-specific adaptation to deal with evolutionarily novel, nonrecurrent problems.
I am not even going to get into what I think of his theory of intelligence. If you’ve read any of my other articles about intelligence, you can probably guess what I would say about that.
In the list of “accumulating evidence” to support this hypothesis, he includes the tendency for less intelligent individuals to respond to TV characters as if they were real friends, and says that people of above-median intelligence do not become more satisfied with their friendships by watching more TV, but less intelligent people do. Huh? He claims this is because TV was not available in the ancestral environment, so that is the reason for the correlation between TV friendships and intelligence, which is supported by his hypothesis.
Hmm. Maybe more intelligent people just spend their time doing other things more academically related, like reading, therefore, they don’t watch as many TV programs, and thus don’t feel as much of a connection to the characters? There are a plethora of better explanations for this relationship (and others like this mentioned in the paper) than the Savanna-IQ Interaction Hypothesis.
Kanazawa also has some interesting, yet flawed definitions of “liberalism” and “conservatism”. According to him, he defines liberals as having a “genuine concern for the welfare of genetically unrelated others” and they have “the willingness to contribute larger proportions of private resources for the welfare of such others.” He then classifies this as evolutionarily novel behavior because humans are designed to help their kin and no others. Apparently conservatives are only concerned with helping themselves, and are not altruistic in the least.
Is he forgetting about the entire religious population that tends to affiliate themselves with the conservative party? They make far more charitable contributions to anonymous people than do any other group. And I know plenty of people who don’t fit into his model of “liberalism”.
He also thinks that atheism is a product of communism (yes, he really wrote that), rather than the more likely explanation that atheism is a product of increased scientific and technological discovery. Things that were once thought to be mystical in nature, such as gravity, are now known to have an explanation in science. The more unknown things that are found to have a root and explanation in science, the less likely people are to believe the mystical explanation. The more we discover about the world through science, the more people question alternate explanations. No Savanna-IQ Interaction Hypothesis needed.
The Savanna Principle and the Savanna-IQ Interaction Hypothesis make no sense given human nature, and should not be used to support his contorted and biased view of religion or political affiliation.
I understand that Kanazawa is trying to come up with an exciting new theory of intelligence here, but he should really focus on basing it in common sense. I think he saw this as a convenient way to fit his theory into a data set a posteriori, and launch it to the public under the guise as a theory proven with solid data.
The link between IQ and religious/political beliefs has been made before in other studies, so he isn’t breaking new ground here with the claim made in the title of the paper. What he is doing is taking longitudinal data and warping the implications of it to suit his unsubstantiated theory of general intelligence.
Andrea Kuszewski, an Affiliate Scholar of the IEET, lives in San Francisco and works as a researcher and manager with VORTEX Research Group. She investigates the neurocognitive factors behind human behavior.
(5) Comments •
(6598) Hits •