It seemed odd that the outliers at the two ends of the spectrum could be detected by what appeared to be unrelated tests. You'd expect that if the founders at one end were distinguished by the presence of quality x, at the other end they'd be distinguished by lack of x.
In general I think it's quite common to need different tests or heuristics for the two ends of a scale. Do you use the same heuristic to detect, among people, the smartest of the smart and the stupidest of the stupid? Or in works of art, or say pop songs -- the bad ones are not "the opposite" of the good ones.
Assuming your expected success as a startup founder can be modeled as a linear combination of various characteristics you have, there is probably going to be a single characteristic with the largest coefficient.
Regarding the smart people example, it might be that a single heuristic that looks for the presence or absence of a specific gene really does do the best job classifying people. Regarding the pop song example, there might be some measure of an artist's musical ability that does the best job. We are looking for the factors that are causative.
In a general population, yes, but for a relatively small sub-population selected for a specific purpose, his assertion seems to hold water.
Not sure if it can be tested, but I have noticed a pattern where the people that I find (subjectively) less likely to update tend to speak in absolutes, rather than qualifiers. For example, "Linux is so much better than Windows in every way" vs. "I like the flexibility Linux gives me, and I saved a few bucks, too."
Paul Graham has a new article out. Everything he's written is worth reading if you're at all interested in startups, but this article seemed explicitly connected to rationality, by identifying an area where people who are more likely to update / less likely to rationalize will do better than others.
The obvious questions: can this be tested? Noticed early on, rather than in hindsight? Changed by rationality training?