We all know that doing better then others is an integral part of our satisfactions. Thus when we think about utility functions, at least in some contexts, it is important to consider not just absolute outcomes but also relative ones. Indeed, there is for example evidence that people are willing to lose something only to hurt (more) others. Presumable this raises their utility.
Klaus Abbink and Benedikt Herrmann push this further with an interesting experiment. They let people chose to pay in order to probabilistically hurt someone else in a game where Nature may hurt the other anyway. And when Nature is in play, people hurt much more each other, and also expect this to happen more. The moral cost is clearly lower, as one can always defend oneself by pretending Nature did it. Strangely, this is even prevalent when there is complete anonymity.