Genome-wide association meta-analysis of 78,308 individuals reveals this shit has gone too far

Everybody is a genius. But if you judge a fish by its ability to climb a tree, it will live its whole life believing that it is stupid.  — Albert Einstein

Mea Culpa

Hello everyone.  In the end, this post is going to make me look like an idiot in some way.  Why?  Because this is what is bound to happen when we speak about intelligence.   In this case I am willing to take this risk, to stick my neck out, because really – somebody’s got to do it.  This shit has gone way too far.

Outline

This post is inspired by a Nature Genetics article, and various media coverage of it.  There’s a whole lot wrong with this article and especially with the coverage of it but I’m going to break it down for you in three parts,:

  1.  Why “General intelligence” of a person is absolutely not a quantitative thing and why it makes us look very silly to pretend it is.
  2.   How correlation is not necessarily causation and how some condition being correlated with a genetic sequence absolutely does not mean that you can call the condition “genetic”.
  3.   How the entire procedure of looking for genetic roots of cleverness is  misguided, so badly so that it reflects very poorly on anyone who might pursue such a study.

So buckle up, here we go.

1)  IQ and g factor and “general intelligence”

To be fair, the authors of the Nature Genetics article in question wisely don’t refer to the IQ in their article.  However, some people do.  Quite intelligent people even fall for this, which is interestingly enough a good anecdote of how general intelligence can fail as a quantitative measure.   You get that?  Let me repeat it: the fact that even really smart people might fall for IQ as being a thing is evidence that IQ is not a thing.

The first glaring error people make is to imagine that there is a characteristic of a person called the IQ.  This is simply not what the word means.  IQ is a way of grading a person’s test score.  That’s all.  We can say “her IQ is high”, and this might be a perfectly fine thing to say in the proper context.  It’s a qualitative statement, there’s nothing wrong with that.  In the same sense you could say “He’s a really big guy” and again there’s nothing wrong with that.  But if you say that his bigness is 115, you probably will get some funny looks.  There are many interesting ways to describe different aspects of intelligence, but is there any use to assigning a scalar number to general intelligence?

General bigness.

The idea behind there being a general intelligence as an attribute came from the fact that people’s test scores on various tests are correlated.  A person who does really well on one test is more likely to do well on another test than a person who totally flunked the first test.  True enough.  And mathematically there is nothing wrong with expressing correlated measurements as a new average quantity plus a peculiar element.  For example in fluid dynamics we might find it advantageous to write the velocity of a molecule as a sum of the average bulk velocity of the fluid plus a peculiar velocity of this particular molecule.  So why not?

Consider then that when we look at various objects and consider their mass and their volume, we will find that they are correlated.  More massive objects are more likely to take up more space.  Why not then consider that there is a general bigness to objects?      We can call it the b factor.  There’s nothing preventing us from writing the mass of an object as some mass due to the general bigness of the object plus or minus a correction.  But what would we gain from such an analysis?   Well nothing, we would gain nothing and would be wasting our time.

Why?  Because we can see that general bigness is not addressing any quantity that helps us to model the world quantitatively.  Volume and mass (and density) are words and concepts that we can apply to every individual situation, not just averages.  General bigness is not such a creature.  General intelligence or g factor is exactly the same, simply not a quantitative object of an individual.

Of course there are smart people, and big objects, and good athletes.  Sure, our abilities to perform some physical activities are also going to be correlated.  Good runners are more likely to be good football players and even are more likely to be good swimmers or rock climbers.  Statistically.  But do we speak of people’s general athletic index a?  No, again it isn’t an immediately useful quantity to consider because it’s not a specific thing.

World Intelligence League

Lets imagine for a second that I’m wrong here, and that there is really a thing about a person, a somehow even slightly measurable thing, called their general intelligence.  In that case it would be possible to perform such a measurement in a rating system and to have a world champion and a ranking, just like is done for ski racing and chess.  You could form a world intelligence league and give people a rating when they join your club (lets call it 1000).  Then you could administer tests and see who did better in the tests, and contestants that had high ratings (2000+) would contribute with a higher weight to the new rankings after the test.  But who would choose the tests?  The problem is once again that there’s no measureable general intelligence to be found.  You also won’t be able to decide a clear “world’s best athlete” nor a clear “world’s best person”.  It’s because these aren’t quantities without a specific context.

If I say “he’s good” what do you think?   Your mind is likely to produce a context from thin air here, and assume I am talking about his behavior towards others, perhaps his ethics, or some such.  However if I follow with “at the free throw line” there’s a very different and clear meaning.  That’s why there’s no general goodness to be measured, no general bigness to be measured, no general athleticism to be measured, and no general intelligence to be measured.  Notice there’s also no general fastness (even though clearly people and machines will have race results of different distances or events correlated).  There’s no quantity we can measure called general healthyness, and there’s no quantity we can meausre called general musicianship.  Oh did I mention there’s also no quantity known as general intelligence?

OK this should be clear as day by now, if I continue it’s likely to be insulting your SAT score.  SAT score, Erdos number, olly height, chess rating, 3×3 rubics cube time – these are things one can quantify.  All of them require intelligence sure.   A person who is good at all these things is generally intelligent.  But pretending this person has a “g factor of X” just makes you look very silly doesn’t it.  Sillyness factor definitely 2.447.  I measured it.

2) It’s genetic 

Oh boy is this tiring.  Repeat after me: correlation is not causation!  Lets give some exapmles:

The world’s top 二胡 players have a genetic correlation present with Han people.  Is playing the 二胡 genetic?

Studies show a propensity of Finnish speakers to be correlated to genes producing low melanin content.  Is speaking Finnish genetic?

Older children have larger feet, and also score better on geometric arrangement tests.  Is geometric arrangement ability due to foot size?

Individuals from families are likely to share similar diets and lifestyles as well as similar genetic material and have similar diseases.  Are cancer and obesity genetic?

Some tests showed the presence of certain genes was correlated with people doing well on some test.  Does that mean the ability to take that test is genetic?

Of course it doesn’t mean that.

One reason people are jumping to such conclusions is that they have been asked a very silly question:  nature vs. nuture.  This is like asking what enabled you to commute to work in the morning: was it the gasoline or the steel frame of the car?  Well it was 80% steel frame and 20% gasoline according to some studies.  Do you believe me here?  Then why would you believe that something like intelligence might be x% hereditary?  The notion is absurd.   You need both gas and steel frame to get your car running, obviously.   Nobody’s going to talk about the great steel vs. gas debate of what powers cars and hopefully nobody will talk about nature vs. nurture anymore either.  If they do it’s because they are 60% lungs, 40% heart.

Don’t get me wrong,  it might very well be interesting for some reason that certain genes correlate with certain test scores.  It’s an observation one could make.  However this observation is not at all, in any sense, evidence that the test taking ability is genetic.

3)  Why the whole thing reflects sadly on those who might pursue this research

Consider for example test scores on predicting numeric sequences.  Sure, we’d like to do better at understanding how to identify patterns and use our imagination to see what underlying concept produces a given sequence.  Why not?   The idea as I understand it is that we are interested to improve ourselves and others – to get better scores.  And to find out how to get still better scores.

Well guess what?  The answer is immediately obvious – practice.  In none of the tests I’ve mentioned here, from cube solving to free throws to numeric sequences, is there any evidence that we are near a genetic limit.  Practicing always improves the score.

It really seems like bringing up genetics in the context of some intelligence tests is only a desperate tactic for those who really are scoring badly.  “It’s just not in my genes”.   Yeah right, in fact the problem is you didn’t practice as much as the other folks.

So quiet down about genetic factors of intelligence, we’re not buying it – instead, go practice a skill.  Take an SAT prep course to improve your score – or you could sit on the couch watching TV and blame your poor score on genetics.  Your choice.

Leave a Reply

Your email address will not be published. Required fields are marked *