On: An astonishing regularity in student learning rate

[SketchySeaBeast]

You can practice IQ tests and get better at them. Did you become more intelligent because your IQ score got better?

[scotty79 → SketchySeaBeast]

I think you can improve your result a little bit, but not much.

[chaxor → scotty79]

I increased by about 20 points after realizing how to apply some of my knowledge of algorithms. That shouldn't happen (measuring knowledge, not IQ). It's quite a meaningless test - probably best understood when trying to apply it to ML systems like Gato or PaLM.

[scotty79 → chaxor]

Did you don't multiple test before and after? 20 points might be just variability for a single test of poor quality.

Raven matrices is the standard.

I notice knowledge of binary operations and computer frame by frame anination helps a bit with some but still a small fraction of questions.

What algorithms did you notice being useful for solving IQ tests?

[lern_too_spel → scotty79]

This goes against what we have seen with the Flynn effect in multiple East Asian countries.

[scotty79 → lern_too_spel]

I don't think that the cause of the Flynn effect is practicing IQ tests a lot.

[lern_too_spel → scotty79]

The point is that you can improve IQ scores a lot by education, some of which helps you do better on IQ tests. Whether it is by practicing IQ tests specifically is not relevant.

[scotty79 → lern_too_spel]

Those effects are not large:

https://www.psychologytoday.com/us/blog/brainstorm/201806/ho...

1 point per year of education is the lower estimate.

[lern_too_spel → scotty79]

15 points over two decades is an entire standard deviation improvement for the whole population. One-on-one tutoring has been shown to make a two standard deviation difference. That's the difference between median intelligence and genius. If that isn't large, what is? https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem

Your study is about the effect of an extra year of badly done education on people who drop out of school, which is worth 1 IQ point. That same study shows that the value of mediocre education on people who don't drop out of school is 5 IQ points.

[scotty79 → lern_too_spel]

That improvement might come from all sources. Health and nutrition improvements can have very high impact. Deworming, using safer water. Contribution of education to this 15 points might be just few points.

> https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem

It's not about IQ, also, was it ever replicated?

[lern_too_spel → scotty79]

> Contribution of education to this 15 points might be just few points.

It certainly isn't genetics. You're not going to cause a population-wide one sigma increase within a single generation using any genetic mechanism.

> It's not about IQ, also, was it ever replicated?

Yes, this has been replicated repeatedly. As far as IQ, if you apply for a private school that uses IQ tests for admission, you will find that all of them explicitly forbid test coaching because it is well known that it works. Even a video will cause large gains. https://www.sciencedirect.com/science/article/abs/pii/S01602...

[giardini → scotty79]

I'm no fan of educating people simply to score higher IQs.

But the effect of education could be very large: e.g., give an illiterate person an IQ test, educate them and repeat. The change will be ~100 points.

To be correct you can improve IQ up to a point (depending on the individual) with education.

[scotty79 → giardini]

No. It can't be 100. IQ tests don't require literacy. They don't even require language. You could administer IQ test to analphabet who doesn't speak you language and if he cooperates he could get result faily close to their best capacity that could be achieved.