I would assume that the exchange of energy between atoms is assumed to be independent in addition to being random. And by independent I mean that how much other atoms have exchanged energy with a particular atom does not have effect in how many atoms are likely to exchange energy with this particular atom. (Or, to be more precise, atom that has received energy from other atoms is not more, but _less_ likely to get energy from next encounters with other atoms)

But that does not hold in society. Preferences of people are highly dependent. If a singer has gotten praise from one person, it is more likely that s/he is getting praise from the next one as well (assuming these can share their preferences), and by more I mean more than what would be expected from singing talent alone. I am quite sure there was a social psychology research on specifically this example, how people rank music in groups, but for some reason I can't find it from google just now.

From the investment point of view, I think Piketty showed that larger investors seem to get better returns on their investments. You can debate causality, correlation and luck here probably endlessly, though.