A note to myself.
An excerpt from Andrew Sullivan:
For my part, I’ve come to doubt the existence of something called “g” or general intelligence, as the research has gathered over the years. I believe IQ is an artificial construct created to predict how well a random person is likely to do in an advanced post-industrial society. And that’s all it is. It certainly shouldn’t be conflated with some Platonic idea of “intelligence.” I don’t think it carries any moral weight at all, either, and I don’t think it should be used in any way in immigration policy. In fact, any public policy that rests on this kind of data is anathema to me. It’s far too close to eugenics, and to the morally repugnant idea that smarter people are somehow better in any meaningful sense. [Emphasis.]
I do not know what Sullivan’s previous mental model was concerning the scientific construct called general intelligence, so I find it difficult to assess this statement of his. For example, had I once believed that g as a psychometric unity, as a common factor extracted from a positive manifold of correlations, had its source (i.e., the reason for the positive psychometric manifold) in a very reductionistic physiological unity, in a singular factor of the brain e.g., dendritic efficiency, I might now “doubt the existence of something called “g” on the physiological level. In the same way, it might be reasonable for Sullivan to doubt g. Basically, I am not certain about Sullivan’s implicit model, so I can not be sure that I am well characterizing his view.
This noted, the impression that I get from the excerpted passage is that Sullivan has regressed towards the mean in terms of his understanding of the issue. He makes the following statements, which might have been defendable two decades ago: The existence of g is dubious; IQ is an artificial construct; IQ predicts performance in advanced industrial societies and only that; IQ should not be conflated with “intelligence”. The latter three statements actually follow somewhat from the first. The adroit statistician David Bartholomew pointed this out in his very readable book, Measuring Intelligence: Facts and Fallacies (PDF). With regards to the first statement, we must distinguish between IQ as a manifest variables — something that can be observed or counted directly — and general intelligence as a latent variables — something that must be inferred and measured indirectly. IQ, which describes scores on IQ tests, is an “artificial construct” in the sense that it is dependent on the composition of the IQ battery given. General intelligence is not an “artificial construct” because, jointly, (a) it is understood as a latent variable which is measured by IQ and (b) that this latent variable is a fact of the world — it is not measure dependent.
As for the first part of Sullivan’s second statement, it’s obviously false, unless you define “performance” to mean simply what IQ scores predict. For example, is “health” “performance”? As for the second part, it’s not clear to me which societies constitute “advanced industrial societies”, but, as Malloy (2008) has pointed out, the predictive validity of IQ tests is non-trivial globally:
Nonetheless, you could construe the statement to mean that IQ tests only measure general intelligence well in modern societies. Is this the case? Probably not, but it’s irrelevant anyways since IQ tests are only measures. One would just have to create a version of IQ tests that measures psychometric general intelligence in the non -advanced industrial society of question. This shouldn’t be impossible, since measures of the g-scores of other primates have been created. We are then brought, almost inevitably, to the third statement. Well, surely IQ measures and IQ scores should not be conflated with “a platonic idea” of intelligence. But that’s not what is happening. Rather, intelligence is being defined as general intelligence e.g., Trzaskowski et al. 2013, the factor common to mental abilities, and general intelligence differences are being measured, with some degree of reliability, by IQ tests. Everyone accepts this practice until the issue of race/ethnicity comes up — and then we all forget everything that has been said on this issue. For example:
Nytimes. January 5. 2012. Intelligence Is Not the Same as Value
People who do well on a test of one mental ability — let’s say a test of verbal ability — will tend to do well on tests of others — math ability, spatial ability, and so on. This finding, which has been replicated thousands of times, implies that there is a general factor of human intelligence. Psychologists call this factor “g.” We still don’t know what underlies g. Ian Deary, a researcher at the University of Edinburgh, has argued that the speed of perceptual processes is one piece of the puzzle, while Randall Engle, of the Georgia Institute of Technology, has established that intelligence is strongly linked to working memory capacity, which he thinks of as the ability to hold information in the focus of attention. Others suggest that when we try to boil down the human intellect to a single factor, we lose view of its complexity….
What we do know is that measures of general intelligence are practically useful. Frank Schmidt, of the University of Iowa, and the late John Hunter, of Michigan State University, documented that g is the single best predictor of job performance across a wide range of occupations — better than personality, interest, motivation and even job experience. People who do well on tests of intelligence tend to make the best mechanics, managers, clerks, salespeople, pilots, detectives and scientists. They also tend to make the best teachers. It makes perfectly good sense, as Andrew Biggs and Jason Richwine argue, to use intelligence as a predictor of teacher performance. We should want smart people to be our teachers.
(But not our compatriots!)
As for a Platonic form, specifically, g as intelligence comes as close as one can. A platonic idea is none other than an atemporal, aspatial entity which makes something what it is. It is a thing’s underlying property or commonality. This is simply a pre-modern conception of a common latent factor. Platonic Idea = Common Latent Factor!
What went wrong with Sullivan? Specifically, he somehow came to the conclusion that the “existence” of psychometric g was still an open question. Possibly he never had a clear understanding of the issue. Generally, it seems that he succumbed to the general idiocy that one find on this topic. He criticizes the strawman arguments of some — i.e., group difference can’t exist because groups are not species — but then passes on his own — i.e., (implicitly) psychometric g is an “artifact”. To paraphrase Sullivan himself: “…But please don’t say truly stupid things like g does not exist or that IQ only predicts performance in advanced industrial societies and only that. Denying empirical reality is not a good thing in any circumstance.”