PDF Ebook Why Don’t You Want To Be Rich?
To explain people’s behavior, we often cite their references. It is commonly accepted that to be explanatory, a preference – in combination with other mental states – must have brought about the behavior in question in the appropriate way. One condition for being an appropriate preference for this purpose is that the preferred alternative stands in a relevant relation to the behavior in question. This restricts the explanatory use of many preferences. For example, an agent’s preference for coffee over brandy at this moment (at 8 am, after waking up) does not explain her choice of coffee over brandy at the end of the dinner party yesterday night. Instead, to explain yesterday’s choice requires a preference over alternatives that stand in some abstracting relationship to yesterday evening’s choices: maybe a preference for coffee over brandy after dinners, or a preference for non-alcoholic beverages. In order to be explanatory useful, the most preferred alternative has to exhibit a degree of abstraction, so that the behavior in question can be related to it.
This condition easily comes into conflict with the need to empirically justify preference ascriptions. Preferences are mental states. Given that introspection does not provide a reliable epistemic basis, preferences cannot be directly observed, but can only be derived indirectly from observed behavior. From choices, however, one can only infer preferences over specific states of the world. For example, observing an agent choosing coffee over brandy for breakfast does not justify attributing a general preference for coffee over brandy to her. Even attributing to her a general preference to have coffee for breakfast rather than brandy would require observing her breakfast behavior under many different circumstances. Choice observations only justify attributing preferences over the very specific circumstances in which the choice was observed.
This presents a problem for explaining behavior with preferences. Only the most specific preferences can be derived from an agent’s observable behavior. But if the outcomes over whichpreferences are defined are very specific, then they cannot be employed in explaining any behavior except for the very choices that justified their attribution in the first place. Such explanations would be trivial. So how can one ascribe preferences that are sufficiently abstract for explanatory purposes in an empirically justified manner?
Posted in :
