The thing we pursue and the thing we feel can come apart. A choice is made before an experience, or after an earlier experience has become a memory. The life it brings must then be lived, moment by moment. Between those occasions lie expectations, recollections and purposes that the act of choosing does not disclose by itself. A preference can tell us which prospect someone selects while leaving open how that prospect will feel, what it will become in memory, and whether the person regards the sacrifice as worthwhile. The question is whether utility revealed in choice measures the good the choice was meant to serve. Answering it requires an account of that good independent enough to allow the choice to disappoint.
Experienced utility supplies one such account by returning attention to pleasure and pain actually undergone. Kahneman, Wakker and Sarin distinguished it from decision utility, the weight of an outcome in a person's choices. They also separated remembered utility, a later evaluation of an episode, from predicted utility, an expectation about what it will be like. These distinctions make disagreement visible. Someone can expect an ordeal to be bearable, suffer through it, remember its ending favorably and elect to repeat it. Each report answers a different question at a different time. Calling the final election a preference records a real decision; the word alone cannot settle which of the other reports it faithfully represents.
Memory can favor an episode that contains more pain. In the cold-water experiment discussed by those authors, a longer trial repeated the unpleasant beginning of a shorter one and added a further interval that was still uncomfortable, though less so. Many participants preferred to repeat the longer trial. The ending changed the later choice even though the extension had added discomfort. That result exposes a possible conflict between remembering an episode and undergoing it. Its reach is also specific: a short, largely passive experience with a reasonably homogeneous hedonic character. A discovery about such episodes gives a reason to investigate a memory-driven choice; it does not establish that every recollection discards duration or that every life can be evaluated as an extended cold-water trial.
A deliberately simple comparison makes the conflicting criteria explicit. Imagine a short episode with pain intensity four for one minute and a long episode with that same minute followed by one more minute at intensity one. Assume that the intensity units are comparable across these minutes and that total pain is their time-weighted sum. The short episode contains four pain-unit minutes; the long one contains five. A rule averaging the worst intensity and the ending, however, gives the short episode four and the long one two and a half. The longer ordeal looks better under the latter rule while containing more pain under the former. The disagreement is produced by two specified ways of evaluating the same history. Whether the total-pain account should govern this decision remains a judgment about what matters in it.
The choice can serve purposes that a pain total leaves out. A difficult conversation may be worth having because it repairs a friendship; a demanding journey may matter because it fulfills a promise. The discomfort is real, and so is the reason for accepting it. Even in a case where the chooser mispredicts the discomfort, that error need not reverse the judgment about the whole undertaking. To identify a welfare failure, one must know enough about the aim and the alternatives to distinguish a mistaken forecast from a sacrifice the person has reason to make. The experiencing life and the remembered story can both matter. Their relative authority cannot be obtained merely by giving one of them the name utility.
Income brings the same problem into the comparison of whole societies. Easterlin's 1974 inquiry separated three relationships often compressed into one: the association between income and happiness among people in a country, the association across countries, and the movement of happiness as a country grows richer. His historical survey collection showed a positive within-country association. The international comparison was uncertain, and the United States time series he examined did not show higher income systematically accompanied by greater reported happiness. He proposed changing reference standards and relative position as an explanation for the contrast. The proposal made the formation of wants part of the problem. It did not establish a universal income threshold beyond which another dollar ceases to matter.
Later evidence changed the terms of that historical dispute. In their 2008 reassessment, Stevenson and Wolfers reported positive relationships between subjective well-being and the logarithm of income across and within countries. They also found evidence consistent with well-being rising as economies grew, while describing the time-series evidence as more suggestive and less precise than the other comparisons. Their analysis resisted a fixed satiation point and retained a possible role for relative income. The resulting lesson is methodological as well as substantive. A relationship among people at one date cannot simply be read as the effect of making an entire society richer, and either relationship can change with the measure of well-being, the countries, the period and the quality of the data.
A more recent disagreement also turned on what the measure could distinguish. Killingsworth, Kahneman and Mellers's 2023 collaboration reanalyzed experience-sampling reports from employed adults in the United States. Averages concealed different patterns within the distribution of emotional well-being: flattening appeared among the least happy respondents, while the positive association with log income continued for happier groups. The study also showed how a response scale near its ceiling could obscure improvements. These were comparisons among respondents, with a defined sample and period, rather than an experiment giving the same people higher incomes. They refine the evidence about reported feeling. They neither settle the national-growth question by themselves nor turn a particular study's dollar range into a permanent boundary of human satisfaction.
A happiness measure earns its use by making its question clear. Asking how a person feels during the day samples an experience. Asking how the person evaluates life as a whole invites a judgment that can include achievements, relationships, aspirations and standards. Watching a purchase reveals a choice from a particular menu at a particular price. Agreement among those observations is informative, and disagreement can be more informative still: it directs attention toward a poor forecast, a changed reference point, a constrained opportunity or a value the other measure has omitted. Combining them requires a theory of the connection. Treating their shared vocabulary as that theory merely conceals the work.
The practical response to a mistaken choice can begin by improving the chooser's evidence. A person deciding whether to repeat an experience may benefit from a record of how it felt, just as a person choosing a job may benefit from learning what its ordinary days involve. Even then, better information leaves room for a decision that accepts discomfort for another purpose. A policy that substitutes an official happiness score for the person's judgment adds a further claim about authority, freedom and the costs of error. The evidence of misprediction can support that debate without settling it. The object is to understand the gap well enough to know what a proposed remedy would actually repair.
What a choice reveals remains answerable to what the chosen life becomes.
Decision utility still measures something worth knowing: how a person orders the prospects made available, under the conditions of the choice. Experienced utility supplies evidence about how some of those prospects feel. Welfare asks a larger question about what makes the person's life go well. A science can learn from all three while keeping the passages between them open to examination. The disappointment of a choice is then a finding with a possible explanation, rather than an impossibility hidden by a definition or a verdict that every choice is mistaken. Happiness gives utility an outcome against which to be tested. It leaves the test's purpose, evidence and authority to be argued. The argument remains open.