Socializing may happen at all levels, for example if the candidate is taken out to lunch. In that context the information may come out. The big problem is that

Socializing may happen at all levels, for example if the candidate is taken out to lunch. In that context the information may come out. The big problem is that the margin of error for judging an individual case on such factors is enormous. If the CEO is a guy and has daughters then there's a 16% chance he will treat his female employees better, or close the wage gap by 3%, than he did before having daughters. If not, then... what? The stat is only applicable to "CEOs that have daughters"... how do you compare with other kids of CEOs? If you are not careful, your brain will start filling in the gaps, creating false correlations, and worse yet, feel encouraged to look for or come up for more, even less substantiated correlations. It's a very slippery slope. Or maybe I'm only saying all this because a long time ago, I thought the fact that my future CEO liked to play the violin meant he cared about the art we were going to produce. You won't believe what happened next! ;)