| OCR Text |
Show Rationality and the Structure of the Self, Volume I: The Humean Conception 159 something like it - rational, even if no one's behavior ever conformed to it. So whatever (T) is, it is not an empirical generalization describing actual choice behavior. This should be kept in mind in what follows. (T) has sufficient structural and intuitive similarity to certain laws of logic - transitivity of implication, for example - that we tend prereflexively to agree with Ramsey in thinking "inconsistent" (in some sense) an agent whose choice behavior violate (T) in a cyclical ordering of F, G, and H: (C) F>G and G>H and H>F (C) seems in some sense inconsistent because we reflexively think, first, that F>G and G>H "imply" F>H; and second, that F>H and H>F are mutually contradictory. But although implication and mutual contradiction are welldefined within the apparatus of classical logic, it is not possible to give them a similarly rigorous meaning here. Thus because H>F does not logically imply not-F>H, F>G, G>H, F>H, and H>F all may be true together. Hence (T) and (C) may be, too. Because Ramsey's value axioms secure (at best) the consistency of value measurement units and value intervals among ranked alternatives and not the mutual consistency of the pairwise comparisons among alternatives on which such a ranking is based, they do not exclude (C); nor, therefore, the sub-logical "inconsistency" that (T) and (C) represent. Whereas (T) is not an empirical generalization over actual behavior, (C) appears to be. As Sen has shown, the theory of revealed preference preserves 29 the distinction between preference and "selection" behavior. This means that intransitive selection behavior described by (C) need not violate the transitive preference ordering described by (T), any more than an agent who sequentially asserts inconsistent beliefs thereby commits a logical impossibility. Just as we save the assumption of rationality in an agent whose speech behavior calls it into question by attributing to her rational beliefs that conform to the laws of logic, we similarly may save the assumption of rational preferences in an agent whose selection behavior is intransitive by attributing to her preferences that conform to (T). Edwards (in "Probability-Preferences in Gambling" (op. cit. Note 23) treats it as involving, at best, material implication. 29 Thus I adopt Ullmann-Margalit and Morgenbesser's nomenclature (see Edna Ullmann-Margalit and Sidney Morgenbesser, "Picking and Choosing," Social Research 44, 4 (Winter 1977), 757-785). But I use "selecting" as the generic term in order not to prejudge the question whether every selection is, in their terminology, a choice (hence reflects a preference), rather than in order to raise the question whether picking is a real possibility. I am convinced by Ullmann-Margalit and Morgenbesser that it is, but my account addresses the canonical preference (-or-indifference) relation in order to preserve the completeness condition on orderings. © Adrian Piper Research Archive Foundation Berlin |
