One of the big lessons Adam Smith has taught countless generations of economists is that homo oeconomicus is all its selfish glory can be quite useful to society. Of course, there are circumstances were selfishness is not the best outcome for society, for example when an activity exerts negative externalities onto others.
In a recent article, Ernst Fehr, Karla Hoff and Mayuresh Kshetramade give an example of a situation where society would be better off if people were selfish: spiteful preferences. This happens when somebody desires to reduce another's payoff only to increase one's relative payoff, and doing so hurts oneself. Society would be better off if this person would just be selfish and not commit such acts of spite.
Two points here: first, we need to define what selfishness is. If it maximizing private utility, then if it is relative outcome that are relevant to one's preferences, then this is what we ought to accept as utility. Where it become problematic for society is if such preferences are widespread and people enter into spite tournaments and mutually hurt each other. But even if there is only one spiteful person, his actions exert a negative externality onto others that needs to be redressed.
Second, how widespread is such behavior? The authors of the article argue it is more widespread than you may think. They conducted experiments in India with a game where cooperation is a Nash equilibrium. They find that between 61 and 73 percent of players punish cooperators, and this is more prevalent among higher castes.
I am saddened by such results, as this is obviously bad news for economic development. If people are so willing to hurt themselves to prevent others to become richer, everyone is going to stay poor.