The article makes no attempt to make any analysis about differences in wealth or education. It is VERY SPECIFICALLY addressing some of Taleb's claims about IQ and is VERY SPECIFIC that that is the goal.
The article makes no attempt to make any analysis about differences in wealth or education. It is VERY SPECIFICALLY addressing some of Taleb's claims about IQ and is VERY SPECIFIC that that is the goal.
There are important caveats to what the author's done here:
1. Taleb's real point is that IQ is mostly useful for avoiding (extreme) incompetence rather than selecting for competence. Taleb constructed his argument using statistics which are either incorrect or underspecified, which invites the kind of analysis the author made. But that also means the author isn't substantially engaging with what Taleb means, because Taleb is abrasive and doesn't care to be specific.
2. You should be skeptical of any statistical analysis which leads with Pearson's r (or Spearman's rho, or whatever correlative measure) or the associated R^2. It is a top line measure which is useful but incomplete. You may miss out on the nonlinear relationship or nonperiodic relationship entirely, either of which could hold most of the variance. There are also many ways in which the variance of two things can be associated without there being a predictive relationship. Simply stated: as a statistician would tell you, if all we needed was R^2 then there would be no papers, only abstracts.
A little bit of statistics can be a dangerous thing. I'm not defending Taleb per se; rather I'm saying that both sides are arguing past each other and we probably shouldn't draw a conclusion from either.
It seems we could put this a little stronger, as the article claims to refute all the quantifiable evidence for the core point. As far as I can tell, it does so.
I also don't understand how your first point works. It contradicts itself. Look, either statistics (because this is a problem with statistics, not with IQ specifically) can predict individuals or not. So either IQ is useful in classifying both success and failure. Or it can't be used for either. Generally using statistical values for predicting individual success or failure is, first, very unreliable in the individual case, and second, not considered "just", for good reason. If you statistically show a group of 10 people has a 90% chance of being criminal, it is considered impossibly hard to pick out the innocent from among them using statistics. Therefore the correct way to respond to statistics is to group people by IQ bands, not to give specific individuals advantages (which, granted, is often done). Of course, this will magnify differences over time, which you may not consider socially just.
Your second argument essentially states that "measures that don't take time into account ... don't take time into account". It is wrong insofar that IQ usually does take time (time since birth) into account, but nothing else. That's not a coincidence, that's because they tried other factors and they didn't matter, but obviously a tautology is always right. The point is wrong because it's useless, not because it's untrue. There are many other factors that could be evaluated against IQ, and there has been a ton of research that did so, and never found much correlation.
And frankly, it boils down to this: "being good at X" correlates across X. That's to a large extent what IQ is. Everybody has seen that incompetence extends beyond any individual skill an individual might or might not have. Research shows that competence does too. It makes sense to some extent: having exceptional skill at one particular thing allows you to advance quicker when it comes to other skills. When those skills are related this is considered "trivially true" (e.g. learning piano to learn violin quicker), but what's harder to see is that this is in reality almost universally true, and this is in fact what our school system turns out to be based upon. Skills you learn in school aren't generally meant to be generally useful, but to allow you to learn "real" skills that can't be effectively learned in school much faster.
No, statistical relationships are not symmetric in general. You shouldn't expect equal absolute predictive power for events two sigma from the mean in opposite directions if the random variable doesn't have a normal distribution. IQ is normally distributed, but many obvious and conventional measures of success are clearly not.
> Your second argument essentially states that "measures that don't take time into account ... don't take time into account".
This isn't what I'm intending my argument to "essentially state". My second argument is just that drawing conclusions from "top line" statistics like R^2 is dangerous and potentially misleading.
Usually cognitive abilities have a step skew to the right. And even IQ fails the sniff test for normal distributions as you can imagine that the difference between 70 and 100 is far greater then the difference between 100 and 130.
I can use that information to predict a wide variety of things with very high confidence, but I can’t use it to tell you how tall any individual human will become. That doesn’t make it wrong nor useless.
Not sure this is the case at all. The conclusion states:
> Conclusion
> All the claims from the article that I looked at, that can be interpreted as something specific and tested in a real data set, turned out not to be correct.
It's true that Taleb's article is a bit of a rant, in his specific style, and that it might not get its quantitative claims precisely right. However, I think that his core points are valid and that the current paper, by focusing on relatively minor quantitative claims, doesn't actually provide relevant counter-arguments.
To be clear, not following the scientific method does not mean your hypothesis must be wrong. But to convince the community at large, that's what you should do.
- Let us return to the point of the correlation test-retest. Unlike measurements of height or wealth, which carry a tiny relative error, many people get yuugely different results for the same IQ test (I mean the same person!)
- National IQ is a Fraud. From engaging participants (who throw buzzwords at you), I realized that the concept has huge variance, enough to be uninformative. [...] Some people use National IQ as a basis for genetic differences: it doesn’t explain the sharp changes in Ireland and Croatia upon European integration, or, in the other direction, the difference between Israeli and U.S. Ashkenazis.
- If, as psychologists show (see figure) MDs and academics tend to have a higher “IQ” that is slightly informative (higher, but on a noisy average), it is largely because to get into schools you need to score on a test similar to “IQ”. The mere presence of such a filter increases the visible mean and lower the visible variance.
> But to convince the community at large, that's what you should do.
I don't agree. I don't think that only scientific (or at least scientific-looking) arguments should be considered valid. Why would that be the case? There are a lot of empirical observations that we can draw useful knowledge from, without resorting to mathematical theories and quantifiable experiments. Not to mention that science itself is far less of a rigorous field than it may seem at first sight.
Because we humans are excellent at convincing ourselves of seeing or believing things that are simple untrue[0]. It turns out that numbers - while they can still be messed with knowingly or unknowingly - are a very useful tool for not only combating your own biases but also convincing others that our conclusions avert these.
Empirical observations are certainly immensely useful - for forming a hypothesis. If the hypothesis can successfully predict certain non-trivial effects, that elevates it over hypotheses that cannot do so.
> I don't think that only scientific (or at least scientific-looking) arguments should be considered valid.
As I said, arguments can be valid without being scientific. But mere human perception is a very fallible thing to build research, results and, eventually, policy on.