This is interesting, but I’m curious how much of the effect is due to getting paid $15, and how much the data was affected by the fact that everyone was isolated in COVID lockdown at the time.
Another potential self-selection bias -- if people know they are signing up to have a conversation with a stranger, perhaps they are already predisposed to be more "pleasant" in conversations, vs a potential curmudgeon who doesn't ever want to speak to anyone, even for money.