A person accepts a wage, buys insurance, pays a toll or keeps a holding for another year, and the transaction leaves a visible record of a choice. The want behind it has to be inferred. Much of the table has proceeded by asking what terms would make that choice worthwhile: the return required to wait, the premium required to bear risk, the price at which one more unit would be bought. The inquiry becomes more demanding when worthwhile comes to mean good for the person who chose. An observed act, a ranking that explains it and a judgment about the life it produces are three different objects. Their connections give utility much of its power. Their possible separation gives this volume its question. A theory can describe a decision accurately while leaving open whether its consequences were understood, whether the chooser sought personal advantage, and whether the result improved the life whose conduct it explained.
A stable account of choice makes claims that can fail, because the same ranking must answer for more than the decision from which it was inferred. Consider three alternatives, A, B and C, available in pairs under otherwise unchanged conditions. Suppose A is uniquely chosen over B, B uniquely over C, and C uniquely over A. If the observed choice is to identify the unique best alternative in each pair, no single transitive ranking can represent all three observations: the first two require A above C, while the third requires C above A. The contradiction disappears if the third choice becomes A, and the ranking A, B, C then explains every pair. This is a genuine restriction. A preference assigned afresh after each act would avoid it by surrendering the stability that made the account testable. So would declaring everything equally good and allowing an unrestricted device to pick among ties; that would explain none of the observed selection through preference. The useful model states what remains fixed and what pattern it excludes.1
The feasible menu also belongs to the fact a choice reveals. A worker who accepts the better of two available jobs has made a comparison within those terms. The acceptance says nothing by itself about a third job the worker could not obtain, or about the wage the worker would choose if bargaining power or the available alternatives changed. A buyer who pays dearly for the only source of water may prefer that purchase to going without and still be much worse off than under another arrangement of supply. These possibilities coexist with entirely consistent preferences. In the same way, a decision made under a mistaken belief can be the best act according to that belief and yield an outcome the chooser would have rejected with better information. The evidence acquires its meaning from the alternatives, the information and the constraints. Calling the selected act best without carrying those conditions forward makes the conclusion larger than the comparison that supports it.2
The chooser's purpose can itself extend beyond personal well-being. Sen made the distinction precise by separating sympathy from commitment. Someone distressed by another's suffering may relieve it partly to remove a distress that enters his own welfare. Someone acting from a commitment may instead choose an act he expects to leave him personally worse off, because he considers keeping a promise or observing a principle reason enough. Imagine returning money that could be kept without discovery, with no expectation that the return will bring compensating pleasure, reputation or future advantage. The return can be deliberate and coherent. A representation of the person's ranking can give it the larger utility number, but that number now orders the reasons governing the act; it does not establish that the return increases the person's anticipated welfare. The moral purpose has been admitted to the account of choice. The account must preserve that purpose if it is later used to say whom the choice benefited.3
Even when the aim is a more pleasant experience, the remembered outcome can guide a different choice from the experience as it unfolded. Kahneman, Wakker and Sarin brought this distinction into their account of decision utility and experienced utility. One experiment they discuss offered a choice between repeating two cold-water trials. Both began with a hand immersed for sixty seconds in water at fourteen degrees Celsius. The longer trial added thirty seconds during which the water gradually warmed to fifteen degrees, still uncomfortable. A majority chose to repeat the longer trial; the milder ending helped it to be remembered more favorably despite the additional period of discomfort. The result concerns a particular experimental comparison, with brief, passive and uniformly unpleasant episodes. It does not establish that every memory follows the same rule or that every choice defeats its purpose. It does establish a reason to measure the experience separately when the claim is that a chosen episode supplied the better experience. The ranking inferred from the repeat choice cannot do both jobs unaided.4
Choosing a measure of benefit therefore requires a judgment about which part of the person's life the measure is meant to respect. Pleasure, satisfaction of an informed preference, the fulfillment of a commitment and the capability to pursue a valued activity can answer different questions. The Hedonimeter sought evidence about experience; The Capability asked what a person was actually able to do and be. Neither inquiry can be replaced by attaching a favorable number to the observed act. Nor does finding a gap between choice and one measure of welfare settle who should be empowered to close it. Freedom to decide has a value of its own, and another person's claimed correction introduces its own information, incentives and power. Kahneman, Wakker and Sarin expressly left that institutional judgment open. The empirical question of whether a choice maximizes experienced well-being and the political question of who may intervene require their own reasons.5
The whole table can consequently keep its account of choice while asking more carefully what each use of that account establishes. A marginal valuation explains a sacrifice under given terms; an expected-utility representation organizes comparisons among risks; a monetary surplus measures gains under specified prices and assumptions. Moving from these achievements to a verdict about whose life improved requires the missing connection to be supplied. The wage, the premium and the toll make that question immediate. The later volume on choice follows the conduct itself through its further complications. Here the quantities turn toward what decisions leave standing. Saving preserves resources, borrowing creates obligations, and repeated exchanges alter the holdings from which future choices will be made. The Stock begins by separating that standing amount from the flow that changes it. Utility has asked what makes an outcome worth choosing; the next volume asks what remains in the person's hands when the choosing has passed.