Rawls’ practice rules...

Rawls’ practice rules, in contrast, are not inductive; they are not the result of such recollection of past events, and they are logically prior to the cases in which they are applied. An example of a practice rule would be the rules involved in games like baseball. Here the rules precede the game. What counts as a ‘run’ in baseball is not the result of looking back at what things have counted as ‘runs’ in past baseball games and then concluding: “well, this must also be a ‘run’”.

Practice rules, rather, give rise to the very possibility that the cases in which they are applied can indeed occur. Thus they are not mere generalizations from past behavior. Practice rules define the very behavior which they at the same time permit. In chess, bishops move diagonally; the issue as to whether or not to move a bishop diagonally is not a genuine dilemma within the context of playing chess.

If someone were to insist on moving his bishop non-diagonally, then he would eo ipso no longer be playing chess. According to Rawls the rules of rule-utilitarianism are precisely practice rules. They are rules which define the very institutions they regulate. The normativity of rule utilitarianism, as Rawls conceives it, is the logical normativity of the system of propositions which describe institutions that rule-utilitarianism itself creates, such as promising and punishment.

The State, for example, does not really have the option of whether or not to punish an innocent person, for punishing the innocent is logically forbidden by the very practice rule which sets up the institution of punishment itself.[^10] Deciding to punish an innocent person is analogous to deciding to move a bishop non-diagonally in chess.

As Rawls would have it: “To engage in a practice, to perform those actions specified by a practice, means to follow the appropriate rules”.[^11] On Rawls’ interpretation, then, the main difference between act- and rule-utilitarianism is not merely related to the issue of where to apply the welfare-maximizing measure (namely, to rules concerned with act-types rather than with act-tokens). Rather, rule-utilitarianism differs from act-utilitarianism in that it is a logical theory.