Saturday, August 27, 2011

Meet Towelie, the IQ Test of the Future

Popular culture warrants serious study. Take, for example, the 1991 movie Boyz N the Hood from the short-lived black-whining-drama genre. The film was almost universally praised despite its predictability and clich├ęs. A major plot point hinged upon the SAT test, which used to stand for Scholastic Aptitude Test before it stood for Scholastic Assessment Test before it officially did not stand for anything. The omniscient, two-dimensional father figure in the movie gave us a pat summary of the SAT and IQ tests. “Most of those tests are culturally biased to begin with. The only part that is universal is the math.” However, data from the College Board disproves his assumption.

Below is the Black-White SAT score gap in standard deviations using College Board data that is available online from 1996 onwards and some additional data that Herrnstein and Murray obtained and published in The Bell Curve.

So, the “universal” math portion of the test produces a larger gap than the “culturally biased” verbal portion. The gap has clearly shrunk since the mid-70’s but is now flat and might even be slightly increasing. This could mean that racial egalitarianism has reached diminishing returns in Black college preparedness. When the pool of participants increases, it tends to create downward pressure on SAT scores because the proportion of people attending college is increasing and adding students who would not have been prepared to go to college in previous eras. To test whether the growth of Black SAT participation is responsible for the lack of decline in the Black-White SAT gap, I added a bar graph for the percentage increase of Black participants minus the percentage increase of White participants.

This shows that the percentage of Black people taking the SAT has been increasing faster than the White percentage, but the size of these increases does not seem to clearly coincide with changes in the gap. Compare this to the gaps between Asians and Whites.

Asians are also taking the SAT at a more accelerated participation rate, but their scores are also improving relative to Whites. They have reached parity with Whites on the verbal and written tests, while their math performance is increasingly surpassing that of White students. Consequently, the notion that cultural bias influences scores would be more logical in support of the conclusion that the strong performance of Asians is not yet at their potential, rather than the conclusion that Black students are suffering from such a bias.

Perhaps in response to such data and the attention brought to it by the bestseller status of The Bell Curve, egalitarian psychologists proposed a concept in 1995 called “stereotype threat,” the idea that internalized stereotypes in test takers undermine performance. My initial reaction to this notion was that the egalitarians were making a desperate reach for an untestable hypothesis for the sake of plausible deniability. However, it appears that this notion has led to an interesting array of research and a number of applications besides the Black-White score gap. They also traded relevance and subtlety to attain testability. After all, implying insults to black students before a test is not exactly standard proctoring procedure.

Now that the evidence for a high heritability for intelligence is stronger with corroborating lines of evidence, it is clear that the future of IQ testing will jettison the test experience altogether. In other words, the future of IQ testing will be examination of its physical underpinnings, including but not limited to genetic tests. A recent “debate” of two YouTube personalities featured an egalitarian who copied the questions from an inquirer speaking with controversial University of Western Ontario scientist JP Rushton. In the interview, the Black Al Jazeera journalist Rageh Omaar asked Rushton, “What then are the genes that determine intelligence? What are they? Can you name them?” The question was probably disingenuous, since it is easy enough to find out that genetic studies show no powerful IQ loci. However, the more up to date reply would be that naming most of them is possible if one has time because a recent study of 600,000 single nucleotide polymorphisms showed that they account for 51% of fluid intelligence. The study was not an exhaustive examination of human genome variation because there are many more differences yet to be studied in the many less common single nucleotide polymorphisms, as well as copy number variants and variable number of tandem repeats. A more detailed understanding could come from studies like the one being led by University of Oregon scientist Steve Hsu, who is developing a genome-wide association study of people with an intelligence level of about three standard deviations above the mean (an IQ of 145) or higher. Tellingly, his study is supported by a third-world country that has a serious need for reform of its own scientific establishment. I would suggest that fear of understanding the genetics of IQ plays a role in the potential for China to overtake America in this area.

Unlike twins studies, genome-wide association studies would advance race realism because probably most IQ variants affect IQ in all humans who have them, and the distribution of these variants in not likely to be geographically uniform.

Therefore, the future of the popular critique of IQ will be less Boyz N the Hood and more Gattaca. In Gattaca, a space flight company surreptitiously obtains genetic information from its employees, making the film a depiction of a horrific dystopia in which unhealthy people are prevented from being astronauts. IQ geneticists need an equally powerful pop culture icon to humanize their research.

After the popping of the dot-com bubble, much speculation and excitement turned to nanotechnology. Drugs would be programmed to target pathology with no fear of side effects. Plates would be programmed to tell us about food content. Computers would be woven into garments. Self-replicating nanobots would perfect manufacturing and become a new type of pollution. Enter Towelie. Towelie first appeared on the program South Park ten years ago. He is an RG-100 Smart Towel designed by Tynacorp with a computer chip and a programmed “TNA” to sense a person’s body moisture, beat the average person at chess, and become a weapon of mass destruction, should he fall into the wrong hands. None of this mattered to the boys of South Park because they were busy trying to find their stolen Okama GameSphere, which Stan’s mom bought for “only $399.99.” (All in all, that is fairly useful advice for the computer industry, considering that it is coming from a cartoon.) I would nominate Towelie to be the mascot for tomorrow’s IQ tests. Towelie clones would isolate DNA from hair or skin cells, decode the genome, run an IQ gene algorithm, and wirelessly transmit an IQ range estimate. All liberal handwringing about IQ validity would be moot. If only we could get Towelie to lay off the weed.


Anonymous said...

Showing three graphs with the only definition being that one side indicates the year and another indicates something indicated by a decimal, presumably the difference in the scores, based on the subject of the graphs, shows nothing. Especially considering that if the axis with the decimal figure represents the difference in scores, its by less than a point, which would determine nothing on a racial level, or even on a social level. Introducing a bar in the graph, that represents the difference in taking the test, with the other side of the graph showing a smaller decimal further proves nothing as the bar is given no explanation as to how it's meant to work, especially since the only new figures introduced with it are in the form of hundredths of a single digit integer. However, I'll give you the benefit of the doubt. Black people are taking the SAT at a rate of less than 10% of an individual in difference less than white people and scoring less than a point lower. What's that meant to prove?

nooffensebut said...

As I said, “Below is the Black-White SAT score gap in standard deviations” (emphasis added). The title of the graph also says “SD.” Expressing differences as standard deviations is useful because it standardizes the scale of the data for each year so that the comparisons are more valid. It is a common statistical tool for making comparisons. Also, it is how the data was given in The Bell Curve. So, to combine the data, I needed to express it in standard deviations.

For example, in 2010 one math subtest standard deviation equaled 116 points. The average White math raw score was 536. For Black people, it was 428. For Asians, it was 591. Therefore, the Black-White difference was 108 points or 0.93 standard deviations higher in Whites. The Asian-White difference was 55 points or 0.47 standard deviations higher in Asians.

When I added the bar graphs, I said that it was the percentage difference, but I expressed it as a decimal point on the right-side axis in the graph. To eliminate this confusion, I have re-posted the blog post with these differences expressed as percentages.

Let me know if this is still not clear.

Chuck11 said...

"For example, in 2010 one math subtest standard deviation equaled 116 points. The average White math raw score was 536. For Black people, it was 428. For Asians, it was 591. Therefore, the Black-White difference was 108 points or 0.93 standard deviations higher in Whites. The Asian-White difference was 55 points or 0.47 standard deviations higher in Asians."


Nice post. I was wondering if increased black participation was masking a decline in the SAT gap. I guess, if so, not by too much. Just as a note, usually when groups are compared, the pooled standard deviation of the groups being compared is used, not the pooled standard deviation of all of the groups. So, for example, the 2010 math gap would usually be said to be: (536-428)/~101.5 = 1.06 SD.

Here are the SAT gaps by year up to 2007: (page: 332)

Chuck11 said...

"Just as a note, usually when groups are compared, the pooled standard deviation..."

Google: Thalheimer, W., & Cook, S. (2002, August). How to calculate effect sizes from published
research articles: A simplified methodology.

And refer to page 4.

Chuck11 said...

"Here are the SAT gaps by yea up to 2007: (page: 332)"

This should be:

Sackett and Shen, 2008. Subgroup differences on Cognitive tests in contests other than personal selection: