How do we determine the well-being of a person when her preferences are not stable across worlds? Suppose, for instance, that you are considering getting married, and that you know that if you get married, you will prefer being unmarried, and that if you stay unmarried, you will prefer being married. The general problem is to find a stable standard of well-being when the standard is set by preferences that are not stable. In this paper, I shall show that the problem is even worse: inconsistency threatens if we accept both that your desires determine what is good for you and that you must prefer what is better for you. After I have introduced a useful toy model and stated the inconsistency argument, I will go on to discuss a couple of unsuccessful theories and see what we can learn from their mistakes. One important lesson is that how you would have felt about a life had you never led it is irrelevant to the question of how good that life is for you. What counts is how you feel about your life when you are actually leading it. Another lesson is that a life can be better for you even if you would not rank it higher, if you were to lead it.

To a practitioner in the social sciences, game theory primarily helps to identify situations in which interdependent decisions are somehow problematic; solutions often require venturing into the social sciences. Game theory is usually about anticipating each other's choices; it can also cope with influencing other's choices. To a social scientist the great contribution of game theory is probably the payoff matrix, an accounting device comparable to the equals sign in algebra.

This paper argues that measurements of opportunity which focus on the contents of a person's opportunity set fail to capture open-ended aspects of opportunity that liberals should value. I propose an alternative conception of ‘opportunity as mutual advantage’ which does not require the explicit specification of opportunity sets, and which rests on an understanding of persons as responsible rather than rational agents. I suggest that issues of distributive fairness are best framed in terms of real income, and that meaningful measurements of real income are possible if opportunity as mutual advantage is ensured.