Garett’s post on the prevalence of sheer malevolence is fascinating, but I’m not convinced.  A key fact about experiments is that many participants just want to please the experimenter.  Once they sit down in the lab, they start asking, “What are we supposed to do?” 

Thus, when an experiment explicitly gives them an option to hurt other players, many subjects take this as sign that the experimenter wants them to hurt other players.  “It’s all part of the game.”  If your only options are to (a) sit still and wait for the experiment to end, or (b) hurt other players, the experimenter is giving subjects a strong hint that they’re supposed to do (b).

I’d be much more impressed by an experiment showing that subjects spontaneously try to hurt others.  Suppose you tell them they can pay some money in order to change others’ endowments.  Start with an example where one player pays money to increase others’ endowments.  Then see if anyone spontaneously tries to pay money to decrease endowments. 

I doubt more than 5% of subjects would do so.  Many people will deliberately hurt each other if an authority gives them a not-so-subtle hint that he wants them to do so.  But few are positively itching to torment a random stranger.