Berkson's paradox is really important to understand

Basically, if you select for two different things in a group, you'll tend to notice a spurious negative correlation between those things in the group

The classic example: if you date people who are either nice or attractive... https://twitter.com/SamuelDataT/status/1355467539602681857

then you'll notice that people you date who are attractive are not nice, and people you date who are nice are not attractive

This has nothing to do with the broader world, just your selection criteria

How does this come up in real life?

If you ONLY hire people who have a college degree OR are super hard working, you'll notice that people who have a college degree at your company don't work as hard

...

If you ONLY apply to companies that do interesting projects OR pay really well, you'll notice the ones that pay really well don't have interesting projects

If you only follow people on Twitter who are funny or smart, the funny won't be smart and the smart won't be funny

It's kind of obvious but once you know this trick, you'll see it EVERYWHERE

https://en.wikipedia.org/wiki/Berkson's_paradox

I was definitely thinking about this looking at the now-infamous Twitter hiring memo

That exact situation came up at a previous job, we went to great lengths to hire from top schools and while they were all great people I enjoyed working with, they weren't superstars or anything

Meanwhile the people who did manage to get in from smaller local colleges, or without a degree, were generally amazing

That's not because of anything to do with schools or types of people or anything

Just Berkson's paradox in action

And it would come up even if the selection criteria was, say, a two-week certification, not a whole college degree