ACT, STATE, AND CONSEQUENCE

According to the procedure Savage (1954) and others have adopted as canonical for representing decision problems, three notions are deployed in the representation: the notion of an act, a state, and a consequence. Many philosophers have followed the lead of Jeffrey (1965) in complaining about a wrong-headed ontology that insists on trinitarianism where monotheism should do. Jeffrey suggests that acts, states, and consequences are all events or propositions.

I do not want to quarrel with Jeffrey's suggestion. To me, something like it should turn out right. But I do not see why it should be supposed, as Jeffrey intimates, that Savage (or, for that matter, Ramsey) would disagree. Perhaps Ramsey and Savage may be convicted of what now seems like loose talk; but it is loose talk easily repaired without damage to the substance of their views. Instead of speaking of acts, states, and consequences, Savage could have spoken of act descriptions, state descriptions, and consequence descriptions, or of act propositions, state propositions, and consequence propositions.

There are, to be sure, important differences in a Savage framework between the attitudes the decision maker has toward act descriptions, state descriptions, and consequence descriptions. State descriptions are objects of personal or credal probability judgments; consequence descriptions are objects of utility judgment and act descriptions of expected utility judgment.

There is nothing in the Savage system to prevent assigning utilities to state descriptions or probabilities to consequence descriptions.