alt Hacker News> Statistical validity != construct validity.
Sure. But in science, we regularly postulate the existence of some construct, and confirm that construct by conducting many empirical tests that return results consistent with the existence of that construct. General intelligence is like that. We can’t see it directly. But we have myriad results that are statistically consistent with its existence.
Results consistent with the existence of a construct are not sufficient evidence for the existence of a construct. We can talk about statistical correlations till the Sun goes down, and I will not dispute that this ethereal 'g-factor' can infer minor to moderate predictions in some domains of people's lives at a population level.
However, I have one question. What evidence is there that this 'g-factor' is actually representative of general intelligence? You may not use the correlation values used to derive the g-factor to support your argument. My understanding is that correlations cannot be used to explain the general factor because the general factor should be what explains the correlations.
If you are interested, I implore you to read this blog from the statistician, Cosma Shalizi, of CMU. His explanation is far better than anything I could attempt to make.
https://bactra.org/weblog/523.html