The univariate mind of the far-right crank.

Normal_Distribution_PDF-2There’s a phenomenon that appears in some of the more conservative parts of  “STEM” professionals/students  which I refer as univariate mind.   Univariate mind is the tendency to abstract  the dynamics of extremely complex phenomena such as  whole economies, the gender wage gap,   dating rituals,  under-representation of certain sexual and racial minorities in industry, poverty, etc. into models that use one or just a couple of variables.  Some of the more committed far-right crackpots extend this ridiculous univariate and simplistic just-so stories into olympian limits, such as white-nationalists connecting the rise and fall of nations,  empires, and modes of productions to a couple of variables such as race, culture, or IQ.

The alt-right are probably the worst culprits of  this univariate thinking.  A superficial  review of some of the intellectual influences of the current alt-right, such as Charles Murray’s “The Bell Curve”, evolutionary psychology, cherry-picked studies from behavioural psychology, and an obsession with IQ reveals this intellectual sickness.   For example, a common justification for white ethno-nationalism is the correlation of IQ with a couple of other variables, such a race,  heritability,  a nation’s wealth and criminality.  Then a simpleminded racist would conclude that because certain races allegedely test lower IQ,  it means they are  genetically predisposed to poverty and criminality, and ergo, policy wise they should be marginalized from positions of power.

However what does one mean by correlation? One can measure the amount of correlation between two quantities using a correlation coefficient. In general,  correlations found in  behavioural psychology, which is probably the number one field univariate reactionaries abuse,  are weak to moderate, in the sense that one can fit a vulgar linear regression of through a data set and find usually a correlation coefficient that hovers from (-) 0.1 to (- )0.8, where 0 means no correlation and (-)1 means that there is a perfect, linear (anti)correlation between two variables.   To give a good idea of what a correlation coefficient means, here are some data sets with  fitted lines and their respective correlation coefficients.

From the above link, it’s evident that a finite (anti)correlation coefficient sometimes is not very impressive,  usually means a weak to moderate trend with fairly large scatters.  There’s also other more complex  metrics  that go beyond correlation coefficients that actually adjust for the number of data points, because a correlation coefficient that uses  two data points is obviously more suspect than one that uses a million data points. The correlation coefficients found in data usually cited by white nationalists and professional misogynists (e.g. coefficients of 0.1 to 0.7) to argue for biological causes of gender and racial disparity are mostly unacceptably low  for the physical sciences, but find their way into the social sciences  because social theories are more uncertain and inexact given that human society is orders of magnitude more multivariate, complex and nonlinear than the electron orbitals of a hydrogen atom.

The lower coefficients, which imply a larger scatter, means that the complex social phenomena that these studies try to model have not only one relevant variable, but many, and sometimes such social phenomena are  not linear and they a can’t be fitted with just a  straight line.  There’s also the question that  over 50 percent of psychology research is non-reproduceable and therefore not trustworthy.  I don’t state this limitation in order  criticize the social sciences by any means, because most of those researchers are aware of the limits, but to warn about  far-right cranks with tiny minds that can only imagine the socio-economic world as a simplistic, linear function that is only dependent  of a few variables (such as IQ, or race). A demonstration of the univariate fallacy is in the book “IQ and the Wealth of Nations“, which is pretty popular amongst far-right “pseudo-statistical” cranks.  The authors made a linear regression between IQ and GDP for various countries and found a correlation coefficient of 0.76.  Not only  did they find a statistical correlation, but made very bold claims about how IQ is a function of these countries’ racial composition.  Not withstanding the poverty of the IQ data itself (for example, the author didn’t have an IQ number for around half of the countries considered and instead interpolated the IQ from  neighbouring countries, also some of the sample size for calculating the IQ in these countries were small and poor),  critics showed  with a rudimentary multivariate  analysis that IQ was much less significant than other factors. In other words, the writers suffered from the “univariate mind” sickness, pushing all sorts of racialist charlatanry based on a one variable linear regression.

Now that I delineated the limits of political arguments based on the univariate fallacy, it would be interesting to explore the question of why these positivistic and vulgar approaches are popular with right wing cranks. One of the most obvious trends worth exploring is the training behind many of the people who peddle these  reactionary beliefs. Although I don’t think most of people involved in STEM are racist or misogynist, there is a significant percentage that are (e.g. Moldbug, James Damore), and I think people like that are susceptible to using the sort of basic statistics and univariate, linear functions one encounters in the typical undergraduate curricula of the “hard” sciences. So these guys (almost always guys) get a small whiff of the power of mathematics and abuse them, without understanding  multivariate analysis, complexity science, chaos, and nonlinear differential equations, which are concepts one learns at the PhD level. This couples with their usual reactionary disdain for the humanities, the latter which eschews the positivistic approach of looking at variables in isolation, and instead deal with society “as given”.   Beyond that, there’s also the larger question of instrumental reason and the division of labour in capitalism, which hyper-specializes humans to the point they are unable to see the world without the filters of their method or trade, reducing the problem of societies into a vulgar line that runs through a  scatter of data.

Far-right charlatans are the most obviously diseased with univariate sickness. However, the illness is more or less generalized at this point. Much of liberal policy, which drives standardized tests, education policy, and our belief in market-based economies, relies on many techniques similar to the one used by racial crackpots.  Namely, the belief that a certain form of linear regression, much of the time using very few variables, is strong evidence for certain policies.  Beyond the problem with this approach as delimited in the above paragraphs, there’s also the problem of dynamism. We live in a class society, with inequalities, war, etc, and  therefore any correlations and laws one could find are only specific to the current exploitative, gendered, racialized and ecocidal social configuration. Ergo the economic and social laws that regulate this society might be meaningless when imagining a world without classes or nations, as Einstein once said in “Why Socialism?“:

But historic tradition is, so to speak, of yesterday; nowhere have we really overcome what Thorstein Veblen called “the predatory phase” of human development. The observable economic facts belong to that phase and even such laws as we can derive from them are not applicable to other phases. Since the real purpose of socialism is precisely to overcome and advance beyond the predatory phase of human development, economic science in its present state can throw little light on the socialist society of the future.

7 thoughts on “The univariate mind of the far-right crank.

  1. “So these guys (almost always guys) get a small whiff of the power of mathematics and abuse them, without understanding multivariate analysis, complexity science, chaos, and nonlinear differential equations, which are concepts one learns at the PhD level.”

    Since when is chaos theory limited to the PhD level and what chaotic dynamic system is it exactly that demonstrates that blacks don’t actually have lower IQs even after controlling for SES?


    1. I am going to pretend you aren’t concern trolling. But the point of the article isn’t questioning the empirical correlation (e.g. race vs IQ), but what are the causal mechanisms. Lots of “rationalist” and “skeptic” types would say it’s “biological” but they have absolutely no way to prove this.


      1. Well the causal mechanism certainly isn’t poverty because, like I said, the effect persists even after controlling for SES.


      2. And, again, what is the relevance of chaos theory here, and why do you think that chaos theory is limited to PhD-level study?


  2. I think you exhausted the pool of synonyms to refer to what you consider far-right crackpots haha

    There’s a way to look at different variables one by one without jumping to conclusions though.
    The correlations between IQ and life outcomes is huge regardless of if you group by race or not.
    Basically the job market “discriminates” by IQ.
    Then you find that mean IQ differ by race. Ok.
    So if nothing about policy is mentionned, who or what exactly is racist : the person stating a fact, the fact itself ?

    You can study IQ, Race and inequalities in order to get to the core of the issue, that doesn’t make you racist, especially if you don’t make any claim policy wise. And in fact by doing that you get closer to finding solutions than anyone who denies facts possibly could.


  3. In a contrary, too often the absence of a simple linear correlation is abused as a proof there is no valid correlation at all: e.g. – the effect of testosterone alone is insufficient to explain all performance differences in sports – ergo, the testosterone effect is not important at all in sports. This is the actual univariate fallacy, IMHO.


  4. Here you are another example:


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s