The other problem with IQ is that it's not a fixed scale, so you can't really compare IQ scores across time. An IQ of 100 is average by definition. Even if the average "intelligence" (or whatever IQ measures, because it doesn't seem to be intelligence as people think of intelligence) rises or falls over time, that average will always be a 100 IQ.
It's not raw, in the sense that it's not an objective measurement. It's a comparison with other humans of the same age that took the same tests. 100 IQ means that you score in a perfectly average way, you're better than 50% of people that took that test and worse than 50% other people that took that test, it's a comparison, not really an absolute score.
So, to compare 100 IQ now with 100 IQ 50 years ago is hard, since you're not using the same test anymore.
There's an effect called the Flynn Effect which is essentially an inflation of IQ, so the tests are changed every few years so that it keeps the same distribution (so that the averagely intelligent human would score 100)
In fact, you can't always compare the IQ tests of 2 humans alive, because the given score is comparing you to the other people of your age, not to the global population. So if you compare the IQ of a kid and middle aged man, it doesn't mean that one is more smart than the other in an absolute way (it's more a theoretical potential)
Let's say, for the sake of argument, that people in the '60s were twice as smart as people now.
The average IQ of the people then would be 100, and the average IQ of people now would also be 100, even though there was a huge difference in intelligence. This is because 100 is defined as being the average rather than being an absolute measure.
Ah. okay yeah that makes sense. I didn't realize it's a relative measurement. I'm surprised it's not more robust. Something like using historical results to compare against, and updating tests in a very standardized way where the math/logic is always fairly similar, but the fact checking/knowledge that requires understanding of current world might be different data wise, but tests similar attributes or qualities.
