Chris Dillow instructs us in his post 'Come Dine With Me: The Economics' that social preferences, or preferences that involve regard for the person with whom one interacts, are pervasive in the British Television show 'Come Dine With Me'. It's a fun hypothesis, and an even more fun forum in which to 'test' it, where a 'test' is a simply discussion of the incentives and the evidence of what people seem to do in several episodes of the show.

The structure of the game is as follows: there are four players, each player hosts a meal, each player rates the other players' meals after they have dined, the player with the highest score wins £1000 at the end.  Initially the game is more like an anonymous interaction because players do not have information about others,  but as it proceeds more information is revealed.  Players rate the host's meal out of ten.  As the goal is to win the game, a rational player should always rate other players' meals as a 0 out of 10 as that would dramatically increase their own chances.  If that occurred, the backward induction solution would be for all players to give zero and no player would win (assuming rationality and common knowledge of rationality). 

But players almost always give ratings higher than zero, which would indicate that either a) they are irrational, or b) there are norms that govern their behavior and their behavior is rational.  Dillow articulates how the correct option is most likely b) and that these players either have preferences over fairness, or, in my view, have a reference point about what constitutes an 'ok' number of points to give and that they add and subtract points from that reference point, rather than use zero as the be-all-and-end-all.  Take a look at the post if you've the time, it's an interesting informal application of social preference theory.

HT: Mark Thoma.