Economics of regret

From Learning with Regret:
"Marchiori and Warglien used neural network models that incorporate "regret" to predict the outcomes of games played by humans. In this context, regret refers to the difference between outcomes attained and the best outcomes that might have been attained if the actor had chosen differently. The models' predictions are based not on conventional, forward-looking expectations of gain, the notion so long at the heart of economic theorizing, but instead on the action propensities that develop through a backward-looking learning process that is driven by regret.

This is an important step in the development of a workable new synthesis. Marchiori and Warglien show that a very simple, parameter-free model can do an excellent job of fitting the long-run tendencies of players in 21 different economic gaming experiments."
Abstract of the Marchiori-Warglien paper here. Unfortunately both the papers need subscription.