The IQ trap: how the study of genetics could transform education
newstatesman.com
newstatesman.com
But by the time the child comes to be educated, it seems that you could get the same information more accurately by directly testing the child's ability. You don't need to test the genetics, because you have access to the actual child. Genetic tests could give us a measure of how well a child will do in life, but in fact we already have this measure just by looking at their schoolwork. The genetics don't add anything.
The article says[1] it can help determine an early intervention or optimal method of teaching before the ability is tested. By the time you test the child, it may be too late.
I'm not agreeing/disagreeing with the article. I'm just trying to extract the part of the article that seems relevant to your question.
[1] half way down the article: “At the moment we are detecting ‘problems’ only when they are visible, and at that point they can be detrimental for the child and hard to treat,” says Smith-Woolley. “Genetics offers the potential for predicting and preventing. For example, from birth we might be able to tell if a child has many genetic variants associated with having dyslexia. So why not intervene straight away, with proven strategies, before a problem emerges?” Whether such a scheme could work for more subtle aspects of intelligence and learning – whether we could realistically and reliably use genes alone to predict them, and then tailor learning strategies to have an impact – remains far from clear.
It sort of argues against the position that genetics determines intelligence though. If intelligence was determined by genetics then you wouldn't think that intervening in a dyslexic child's education a couple of years earlier could make a large difference.
Color blindness is easy to test for, as soon as it matters. We can do the tests where figures pop out with baby chimps right? What would change if you knew this earlier?
Are we even sure that dyslexia is something more than a catch-all term for kids who have more difficulty learning to read? And either way, do we know that starting earlier, or later, would be an effective intervention? The error bars on our knowledge of such things (even for the largest samples, all kids) seem huge.
So, if you like, you can replace dyxlexia with condition A, Italian with subject B, and French with subject C.
But we seem to have difficulty even discerning the best ways to teach normal kids to read. Hence my skepticism that our interventions for something specific could be so well-understood that starting the intervention when we had the genetic test (rather than waiting for the classroom) would matter.
Teaching in the UK at least seems quite faddish. At the moment, pre-teen children are taught extensive vocabulary to talk about how English is constructed.
They are taught terms like "split diagraph", which sound liguistic, but aren't terms actually used by linguists!
But I wonder for instance if we might do a lot better at judging schools, and learning what works. Right now the default position appears to be that the input to schools is uniform and thus that good/bad exam results are a measure of the school. But they aren't. If we become a bit less blank-slate about the children, maybe we can do better at figuring out what kinds of schooling work better or worse. This doesn't directly need genetic tests of anyone, but it does need a change of mindset.
This would be a very bad form of evaluation as evidenced by people with unconventional backgrounds becoming successful like Steve Jobs and some countless other smaller business owners who is say show a different type of intelligence.
For sure there are people who seem to take to subjects, to have a natural ability at it; for instance, I took to math. But hard work plays a very big role, a critical role, as well as things such as ability to focus, or whether you struggle with a problem such as an addiction or have bad luck and get in an accident or come down with a disease.
Intelligence, whatever that is, is necessary but nowhere near sufficient to succeed. We do the best we can when we help people work on what they can change rather than telling them they have a certain amount of g and that's it.
Blaming genetics for the results of collective political and economic underachievement is one stop short of blaming individuals for "poor breeding."
Now it is easy to measure GPS and given that it will get easier in the future we'll probably see less standardized testing. I wonder if in optimizing for GPS we might fall afoul of Goodhart's Law.
Someone things worries me is, if we tell children that 50% of the probability of their educational success is "determined" by their genes they will behave differently.
Students who believe their intelligence could be developed (growth mindset) out performed students who believe their intelligence is fixed (fixed mindset). Will focus on heritable intelligence shift more people towards the fixed mind set?
https://www.edweek.org/ew/articles/2015/09/23/carol-dweck-re...
Here's the full process:
-Sequence the genome of everyone
-Identify eugenics supporters
-Select a set of GPS parameters such as the eugenics supporters score below the threshold
-Eugenics supporters are now sterilized
-Problem solved (after all it's just PC and a liberal deception to say eugenics support doesn't have a genetic origin, man isn't tabula rasa after all, eh?)
In similar vein: plan on having the morally handicapped as the admins of the project... that'll not only make it viable, but also will take them out of other projects that need moral leaders.
Sure, I'm wary of the determinism, but I'm more wary of the prejudgement of a person based on their ethnicity, and the scientific pretense that could be stacked onto that prejudice.
> They surveyed almost 2,000 primary school teachers and parents about their perceptions of genetic influence on a number of traits, including intelligence, and found that on the whole, both teachers and parents rated genetics as being just as important as the environment
This is precisely the kind of "begging the question" that makes me doubt the whole enterprise. What do grade school teachers know of the scientific literature? Isn't it more likely that they just see a minority kid in the class underperforming and think, "well, that's minorities for you"? And isn't it possible that a more rarefied form of this commonsensical stupidity is getting fed into such purely statistical pursuits as genetic research?
Intelligence is pretty much always better unlike stuff like height or agreeableness. It's governed by a huge number of genes in complex ways. Humans went through a genetic choke-point not too long ago and we're very genetically homogeneous compared to most species. There are some important genetic racial difference but they all seem to be simple stuff involving single proteins, melanin production levels being a prime example.
Or last week's news about adaptations to diving -- one tiny niche which happens to be relatively easy to study:
And I wouldn't be surprised that some environmental adaptations might be better than others given that relatively little evolutionary time has passed and the barriers for the transfer of altitude adaptations from one high altitude region to another. But there are no such barriers to the transfer of IQ enhancing alleles, those will help you everywhere.
The point of the Tibetan story is that they (may) benefit from > 200k years of adaptation to that geography, via these ancestral humans. Whereas the Andean people had < 10k years.
Re transfer of IQ, why is it always beneficial? There are always trade-offs -- more blood to the brain is going to make it harder to survive famines, for instance. Some monkey species are dumber than others because it was not profitable in their niche. Domesticated herbivores are dumb too. It could be that this doesn't apply to us, e.g. if we are the result of some recent breakthrough.
The flores island hobbits are another example, where it seems that humaniods evolved quite dramatically smaller brains. But islands are also unusual, obviously.
I suspect that sexual selection and social class will become bigger elephants in the room than ethnicity in the future.
Traditional class societies had genetic bottleneck between social classes and ethnicity. There was small flow from small upper classes to lower classes (bastards) but upward selection was restricted for a long time. Smart, healthy and beautiful peasant was likely to stay a peasant and marry from the neighborhood. Stupid ugly aristocrat was likely to stay a aristocrat and marry someone from his class. People were not genetically that different. If there was genetic disadvantage it was with higher classes.
In the last 100 years there has been more movement between classes and people move and travel more. Young educated people marry people around the world.
There is a real fear that upper social class will have real gene advantage that makes them smarter, beautiful and healthier and lower classes will have relative disadvantage.
The effect can also be nonlinear.
As far as I know correlation between wide polygenic score (GPS) and intelligence is linear (r ≈ 0.30) only because linear correlation is measured? It would be interesting to see the effect of wide polygenic score to different extremes, very high and low g.
I would hypothesize that genes are more important factor in low intelligence scores than they are in high IQ. Environmental factors seem to work the same way.
There is some evidence that there is a right skew to the IQ distribution and it's not perfectly normal.
I tried to avoid using the 'IQ' and use g-factor or intelligence, but I slipped at the end. I'm sorry about that.
>. So the high intelligence tail is heavy
Yes. Intelligence seems to have fat tails on both directions and it's also right skewed.
This doesn't need much explaining, the correlation between the score and test results just isn't that high. That's all. Saying "environmental factors" here is, I think, nothing more than a polite way of saying "noise".
People think about nonlinearities too. My understanding is that we're a long way from having such good data that it would help our predictions to include them. Steve Hsu writes about this, I can't find the perfect link but here's a start:
http://infoproc.blogspot.com/2014/11/additivity-and-complex-...
For one reason IQ tests are typically convergent in that they typically test a central way of doing things against a single most correct answer that often excludes creativity. A divergent test, however, examines creativity by looking for the quantity and diversity of answers to questions by a participant.
Secondly, IQ is mostly not an earned quality. Intelligence is generally an inherited quality, but it can be modified by practicing a specific skill that is directly tested for in IQ tests. The ability to improve IQ by practicing a skill is a measure of behavior rather than intelligence.
A far more important quality is the behavior by which skills are learned. Qualities like persistence, determination, focus, criticality, objectivity, and so forth do more to influence perceptions of intelligence than there mere speed of learning, which is what intelligence actually is. That said measurements of speed of learning or raw information processing are weak indicators of potential to excel where those qualities are important.
We have communities of many types that prioritize other things (money, hard work, toughness, etc.) above education. We need to educate people that a good education is the cornerstone of success, and should be prioritized perhaps second, after health. I think we shouldn't expect real results for 3 generations. This is a long term problem that will require major cultural shift, and thus a long term solution.
Defining the concept of 'a good education' is a complex topic. I don't think it should have to include colleges though. My dad was a self-taught successful writer, who didn't go to college. I'm a successful software engineer with only two years of college (I'm defining successful in my case as job I like, working with good people, and a decent 6-figure salary). I am not an expert on education, but I have three ideas I think might work if done together.
First, a massive investment in free basic computers in every house and free online education.
Second, a resurrection of the master-apprentice system, not just for jobs like electricians but for all jobs.
Third, an overhaul of colleges to make them state-run, not-profitable, and ideally free. If necessary you could charge a same cost fee for all colleges of something like 10% of your income for 10 years after college. Or, perhaps, colleges could make money off of research performed there, though that might be a rabbit hole not to go down.
These changes though would require an immense shift in the way we perceive education and being educated, and that is going to be a steep uphill battle.
"Psychologist László Polgár theorized that any child could become a genius in a chosen field with early training. As an experiment, he trained his daughters in chess from age 4. All three went on to become chess prodigies, and the youngest, Judit, is considered the best female player in history."
You'll find a lot more high-iq than low-iq people at top jobs, for example. I'd be very surprised to meet a google/facebook/amazon/etc senior developer with IQ of 85 or less; and while those companies are big enough that maybe some such people exist, it's immediately obvious that iq correlates with success at such jobs.
People tend not to understand weak or loose correlations, and they will assume that race, gender, ethnicity or other factors strongly predict IQ, and that, therefore, systems of prejudice built around those factors can be justified as an expression of the natural order.
Whereas really if genetics strongly predicted IQ, it would be a great argument for a bigger support net. If we believe in equal opportunity, and some people were denied equal opportunity due to genetics, it makes sense to help them. You can no longer say that they're just 'lazy', or whatever.
I don't think it says that, it says that IQ determines the level you could likely get to in an intellectually rigorous occupation. i.e. Top, world-renowned surgeons have a higher IQ than mediocre surgeons, top lawyers tend to have a higher IQ than mediocre lawyers etc.
For example if you look at NBA players, you'll see that they're on average quite tall, but also that height only has a weak correlation with 'basketball ability' in that sample. That's because those people were already heavily selected on height; and all the not-that-tall people in the sample compensate with the other things that make them good at basketball.
So if you want to look at how useful IQ is, you should be looking at the average IQ of successful people vs average IQ of unsuccessful people, and not how successful people are in a relatively high IQ subsample (i.e. all surgeons).
Which of the following do you think are not 1. being smart 2. related to IQ:
1. Ability to reason out answers to questions or problems 2. Ability to make plans 3. .. learn 4. .. generalize - build abstract models from limited data
These are all very much IQ things that people generally call being "smart", and which very much correlates with success in life.
If you need to bet on whether the average IQ of ten random physicians or ten random janitors is higher, which way would you bet?
However, the article is extremely politically charged and filled with identity politics that should have no place on this website.
Only true in some cultures or over some very high range of IQ. Income correlates highly with IQ from a normal low to a normal high range and income correlates with reproductive success in male.
https://skeptics.stackexchange.com/questions/5850/intelligen...
Please don't believe that IQ related studies are mainstream in social and psychological literature: they are very contested. That said, citizens and policy makers should decide if trusting them or not.
There's lots of interesting scientific work about definable attributes of thinking beings - like problem-solving ability, or memory, or spatial sense. These don't suffer the same 'politics' problem, since their topics are clear and possible to concisely express.
Think of this number as the first principal component, generated from all these different things. Talking about it doesn't imply that other directions don't exist. But it's not so strange to work on the leading effect before worrying about sub-leading ones.
But more seriously, yes, to figure out how useful the first component is, you want to know something about how much bigger the first eigenvalue is than the second.
For the case of IQ, my understanding is that the answer is "quite a lot". The best studies are from the army testing people and assigning them to roles (and seeing how they do). And it seems pretty hard to come up with tests which are better predictors than IQ. I think the one perpendicular component worth measuring for some jobs is manual dexterity? And eyesight, which is a rare thing anti-correlated with IQ, because of teenage bookworms we think. (That said, note that this isn't the whole population, they set a floor by only accepting reasonably healthy young men etc.)
I don't doubt that IQ, like Social Credit, is a useful administrative tool. However, something being useful from an administrative standpoint does not make it scientific. It also doesn't make it desirable. There are very many things that are absolute inconveniences from an administrative standpoint, that are absolutely necessary from a human one. Rights, basic freedoms, and so on.
This is a great idea, morally: "We hold these truths to be self-evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights". But as you know if you've ever watched the Olympics on TV, "created equal" here isn't a scientific fact.
If you look at all sports you'll find the same thing. We call the common factor physical fitness, or athleticism, or let's say "AQ". It's exactly the same idea. AQ predicts running speed, although not nearly as well as getting the kids out on the track. Ditto weightlifting. People know this: the kids in the chess club won't agree to arm-wrestle the kids on the track team, because despite having no direct information of their biceps strength, they can guess based on overall impressions. This strategy for keeping their lunch money is scientifically valid. And their profiling of the jocks is a safe distance from being morally repugnant.
If make statistics about ill-defined, politicized questions, at best, you get nonsense. At worst, you get a reinforcement of your muddy, politicized thinking.
The problem with science is, if you're not very careful, it can become a wrapper around stuff that isn't scientific. The only way to avoid this is to be absolutely allergic to any concepts that are unclear, uncertain imports from the normal world. And, while I'd love it if intelligence wasn't like that - let's face it, it is. If it wasn't, and we had a clear image of it, we'd probably be capable of doing stuff like building general AI.
Do you think we understand muscles at the level of detail we'd need to understand neurons to build AI? They are much simpler, but there's a lot of crazy stuff about how myosin works we don't understand. And it doesn't matter, Victorians with stopwatches could correctly gather data about runners & swimmers & wrestlers, and notice what predicts what. Science at this level can be valid too.
And what I'm trying to tell you is that intelligence research is not all bunk as you imagine. There is lots of data. Lots of people have tried to shoot holes in it, for a century, and we keep the bits which survived that. Yes we use words which had everyday meanings before, because we always do that... I mean "energy" and "power" still have wooly everyday meanings, but are used precisely on your electricity bill.
It's like if a medieval scientist set out to study the cosmic spheres. He could amass as much data as he liked, but it doesn't matter - since he's fundamentally using the wrong concepts to understand what he's seeing.
Not to mention, I think your analogy works against your point. If you think about energy or power in any other way than the specific electrical sense, you're making a mistake. Just because they use the same words doesn't mean they are the same concept.
Temperature is another good example. Most of the objections you are raising were raised against this single number when it was being invented. How can the many-faceted feelings of weather become a single number, etc. Your number doesn't fit my pre-existing idea, you're studying the wrong concept, etc. All the same squirming. Lots of messy back and forth. But what survived all that was a number which turned out to be scientifically useful for predicting stuff, like how well your steam engine would work. So it got accepted. (Much later we learned what was going on microscopically, but this wasn't necessary.)
What we're talking about here is essentially epistemology. What is science, and what is its epistemic status?
My feeling is that to the extent that scientists are capable of coming up with good questions, they have cast off the baggage of unscientific and pre-scientific thought systems.
The problem is, it is not the case that a hypothesis is either good, or bad. Mendel proved that even with a very bad hypothesis, you can do very good science (he was a monk). A society of scientists who were incapable of coming up with good hypothesis, whose experiments always proved the null hypothesis - would not be too bad at science.
A society of scientists who were very bad at formulating, identifying, and outlining problems, on the other hand, would be very bad at science. They could devise experiments where neither the null, nor the alternative hypothesis, were of any value in regards to the world.
You can come up with any number of questions like this. The most famous one is probably, 'when did you stop beating your wife?' The most scientifically relevant one is the search for the philosophers stone. You can look through the history of science and see a ton of these questions - areas where vast amounts of effort were essentially wasted, where the only useful results were side effects.
They are universally marked by one trait: they attempt to solve questions that are obvious and important to a non-scientist, and, in terms that have a root outside of the sciences.
Really Interesting science problems are, on the other hand, almost invariably about topics that non-scientists wouldn't find interesting. Feymann spent years working on wobbling plates - and won a Nobel prize for the application of this research to particle physics!
So, if you find yourself trying to answer non-scientific questions in non-scientific terms, the question is, can your answer be scientific? You're essentially saying yes: enough research into the philosopher's stone would make the philosopher's stone science. I'm saying no: the only science that could come about from searching for the philosopher's stone would be an accident. The philosophers stone, unless it divested itself entirely from its non-scientific content (as in the case of energy, temperature, etc), will never be science.
PS: I think there's fairly obviously a conceptual distinction between the concept of power, as in, watts, and the concept of power, as in, Donald Trump. A similar distinction would be between 'vectors' and 'arrows'. I think such conflations are actually a major and frequently encountered obstacle to teaching. But you're welcome to disagree.
IQ is similar. Lots of flaws, but how else do you measure intelligence? There are lots of correlations with IQ. Not using intelligence as a measure is silly.
I think it’s just something that uncomfortable, so I like to pretend it doesn’t matter.
If I’m starting a software company, it is useful to have a measure of raw intelligent. Not to he exclusion of other characteristics, and maybe not even the most important. But certainly useful data for planning.
I do think it’s silly to say the totality of intelligence is represented in a single number. But it can be useful. Like credit score. Lots of flaws, but what is a better measure?
https://www.sciencedaily.com/releases/2012/03/120329142035.h...