Choosing, wanting and being well are three things the theory has often asked one word to hold. A person chooses an option; the choice is taken to reveal a preference; the satisfaction of that preference is then taken to serve her good. Each step has a use, and each adds something that the preceding step alone cannot establish. The volume on The Utility began with the wanting that gives things value. This volume rebuilt the inquiry around the decisions that can be seen. Its last question concerns the distance between those foundations. We can observe a purchase without seeing the satisfaction it brings, infer a ranking without knowing why it matters, and respect a decision without claiming that it maximizes the chooser's happiness. Keeping these distinctions is how the science of choice remains useful when it is asked to speak about a life.1
A stable ranking gives the theory genuine content because it rules some patterns out. Suppose a chooser consistently selects A alone from A and B, and B alone from B and C. A single complete and transitive ranking that explains these unique choices must place A above B and B above C; it therefore places A above C. If the chooser instead selects C alone when A is also available, that pattern defeats the proposed explanation under its fixed conditions. Calling the selected option best separately on each occasion would evade the test. The force comes from requiring one ordering to explain several decisions, with the relevant circumstances held constant. Learning, changing prices or a change in the goods themselves may alter the comparison, but those changes need an account. A model with specified restrictions can succeed or fail without becoming a portrait of everything a person values.2
Even a perfectly consistent ordering leaves open what the chooser is trying to achieve. A person may take a costly turn caring for a relative because she believes it is hers to take, while expecting less comfort than she would obtain by leaving. Sen called attention to commitment: a decision can express a principle whose claim the person accepts despite a loss in her anticipated personal welfare. Concern for another that directly improves her own well-being is a different case, which he treated as sympathy. Both can enter coherent behavior. Assigning a higher utility number to the chosen act may represent the decision accurately, but that number does not establish that the person expects to feel better. To respect her commitment is to take her purpose seriously; to redescribe every sacrifice as a personal benefit would remove the distinction she herself may be making.3
The alternatives that were actually available also limit what a decision can reveal about wanting. Suppose a person has eight, an option A costs ten and an option B costs six. Her purchase of B establishes no revealed comparison in favor of B over the unaffordable A. If her resources later rise to twelve and she buys A at the same prices, one stable ranking can explain both decisions. The change supplies no evidence by itself that her tastes were manufactured or that her earlier choice was irrational. It shows why the conditions surrounding a decision belong inside the inference. The same discipline applies when information is missing or an option carries a consequence the observer has omitted. A record of what people obtained must preserve the limits under which they obtained it before it can become an account of what they preferred.4
Conflicting choices can leave some welfare comparisons unresolved while preserving others. Bernheim and Rangel developed an approach that takes guidance from alternatives unambiguously chosen over others within a specified domain. Imagine three options and two ways of presenting each menu. Whenever A or C is available with B, the chooser never selects B. Between A and C, however, she selects A under one presentation and C under the other; the same reversal occurs when all three are offered. The record supports setting B below A and below C under the approach, while leaving the comparison between A and C unresolved. It preserves coherent guidance without inventing an answer where the choices conflict. If B were itself chosen in the presence of A, the unambiguous comparison favoring A would fail. The criterion has a boundary that evidence can cross.5
The selection of the circumstances that count for welfare remains a judgment the analyst must expose. Choosing with a clear account of consequences can differ from choosing under a misleading description. A person can endorse a course after reflection and abandon it under immediate pressure. Treating one decision as better guidance may be defensible, but the reason cannot simply be that it delivers the answer the institution wanted. Bernheim and Rangel make this choice of domain explicit: keeping conflicting observations can preserve ambiguity; excluding some requires an account of why they should carry less normative weight. Experience, explanation and knowledge of the decision's conditions can inform that account. None supplies an automatic license to replace another person's aims. The remedy for a doubtful inference from choice is a better-supported judgment, with the remaining uncertainty visible.
The want is therefore neither fully disclosed by the choosing nor made irrelevant when the disclosure is imperfect. Choices remain evidence of what people will do under particular conditions, and often provide the strongest available guidance about what they seek. Consistency gives that evidence structure. Attention to feasible alternatives gives it context. A judgment about well-being asks something further: which consequences matter, whose purposes are being respected, and why the selected outcome serves them. The table can describe the choosing with precision while leaving that judgment open. Its useful discipline is to keep the steps distinct, so that a preference inferred for prediction does not quietly become a certificate of a person's good. The next volume follows the recurring departures and biases in the act of choosing. The older question travels with it: what kind of wanting is being served?