Foundationalism About Wellbeing Measurement: A Defense

Abstract

Foundationalism about wellbeing maintains that we can only know what the fundamental constituents of wellbeing are through philosophical reflection, but that knowledge about the causes and correlates of wellbeing measures are irrelevant to this pursuit. Recently, this view has been criticized, and an alternative has been proposed: coherentism. I first argue that coherentism faces a fundamental problem in the context of wellbeing measurement: it cannot function as desired without committing the naturalistic fallacy. I then reject the three arguments against foundationalism: none of these arguments provide us with a good reason to reject foundationalism.

1. Introduction

Wellbeing is increasingly studied across a broad range of social sciences, such as economics, psychology, and sociology, and specifically in emerging subfields specifically devoted to the topic, such as happiness economics and positive psychology (see Di Tella and MacCulloch 2006; Dolan et al. 2008; MacKerron 2012 for literature overviews about happiness in economics; Diener et al. 2009; Diener 2000; Seligman and Csikszentmihalyi 2000 [2014] for (positive) psychology). The study of wellbeing necessarily involves conceptual questions. Imagine you want to know how much an additional hour of commuting time will impact a worker’s wellbeing. The wide range of wellbeing research offers you multiple measures to choose from, such as measures based on people’s self-reported life-satisfaction, evaluations of their happiness, their momentary emotions, of their preferences, or their capabilities. These measures do not result in the same conclusion about the badness of commuting times. How can we determine which of these measures captures wellbeing best?

This question touches upon two important epistemic questions. A first is about the fundamental constituents of wellbeing: which goods non-derivatively make our lives go well? A second is about the measurement of wellbeing: how can we best measure wellbeing? Or, alternatively, how do we know whether a purported measure of wellbeing is indeed about wellbeing? In this article, I focus on two prominent approaches to this latter question. A first view, foundationalism about wellbeing measurement (foundationalism, for short), ¹ maintains that we can only know how to measure wellbeing, or know whether a measure indeed is about wellbeing, once we know what the fundamental constituents of wellbeing are. In other words, according to foundationalism, philosophical theories of wellbeing are prior to measurement. We can only justify our measurements on the basis of philosophical theories and the philosophical analysis that supports them. On a second, alternative view, the two questions are importantly related. Coherentism (Hersch 2022) maintains that we can, and should, learn simultaneously what the fundamental constituents of wellbeing are, and how to measure it, by mutually adjusting both measurement constructs and philosophical theories of wellbeing in order to make them cohere. ² Coherentism builds on examples from the natural sciences, such as the measurement of temperature (Chang 2004), where through empirically investigating a phenomenon, we not only learned to measure it, but we also adapted our understanding of the nature of the concept. Coherentism maintains that to develop good measures of wellbeing not only requires us to shape our measures to what we know about the fundamental constituents of wellbeing through philosophical theories, but also requires us to adjust our philosophical conceptions of wellbeing to fit better with empirical findings.

My aim in this article is to defend foundationalism in the context of the study of wellbeing. The crucial difference between these two approaches is that foundationalism maintains that while our understanding of the fundamental constituents of wellbeing has significant implications for the measurement of wellbeing, the reverse is not true. Foundationalism has recently been criticized by Michael Bishop (2015) ³ and Gil Hersch (2022), who argue it fails because (1) philosophers do not agree on what wellbeing is, (2) that it leaves an unbridgeable gap between wellbeing theories and measurements, and (3) that it puts too much faith in standard philosophical methodology. ⁴

Coherentism has also recently been criticized by Antonin Broi (2025, who calls it the “coordination approach”). Broi argues that the coherentist method requires a target concept that has “open functional content.” A minimal understanding of wellbeing that some have proposed (Hausman 2011; Taylor 2015) is too limited to get the method off the ground, while more substantive definitions are no longer functionally open. While my argument is different, it aligns with Broi’s in that claiming that a coherentist method cannot get off the ground without making substantive commitments about wellbeing. Roberto Fumagalli (2022) has recently also argued that we may find a novel approach that takes the strongest features of both coherentism and foundationalism. Finally, Lange and Grünbaum (2023) have argued that the difference between the two camps (the foundationalist and the coherentist) is due to a more fundamental methodological commitment. Coherentism relies on naturalism—the view that the theorizing about wellbeing should follow the same procedure as the study of natural phenomena—while foundationalism relies on non-naturalism—the view that theorizing about wellbeing fundamentally has “a strong evaluative dimension that should be decided by the methods found in (moral) philosophy” (Lange and Grünbaum 2023, 950). They argue that methodological procedures endorsed by coherentism will have little traction, unless you are already on board with methodological naturalism. I think they are right, and the argument below is intended as a defense of the non-naturalistic stance.

I will first discuss the problem of the evaluation of wellbeing measures in more detail in section 2. Then, I will present the rationale for coherentism (section 3), and discuss my argument against coherentism (in favor of foundationalism) in section 4. I discuss and reject the three arguments that have been raised against foundationalism in section 5. Section 6 concludes.

2. The Problem

2.1. What is Wellbeing? The Very Concept

The debate about the appropriate approach to wellbeing measurement intersects with a debate about the semantics of wellbeing (Alexandrova 2012b, 2013, 2017; see also Fumagalli 2022). Philosophers of wellbeing widely agree upon the semantics of the notion of wellbeing: it describes how good people’s lives are for those living them (Brülde 2007; Sumner 1996; Tiberius 2006). If someone’s wellbeing is high (low), this means that their life is good (not good) for them. This meaning provides wellbeing with its significance: it matters how good people’s lives are for them. ⁵ Alexandrova (2013) has recently argued that this usage does not correspond to many of our everyday usages of the term: doing well at the doctor’s office has a different meaning from the notion of “welfare” in a discussion about economics, or in the context of looking back on one’s life from one’s deathbed. She makes a case for contextualism, according to which we should accept that “the semantic content of sentences in which ‘well-being’ and its cognates occur depends on the context in which it is uttered” (Alexandrova 2013, 310). In other words: “A developmental economist just means something different by ‘well-being’ than does a clinical psychologist. Perhaps the most important consequence of this view is the impossibility” (Alexandrova 2013, 310). The philosopher’s usage of the term—which Alexandrova calls “wellbeing all things considered,” and Hersch (2022) calls “wellbeing simpliciter”—is technical, and different from the meaning of wellbeing in these different contexts (though see Fletcher 2019; Fumagalli 2022 for a strong argument against contextualism about well-being discourse).

We can accept that people do use the term in different ways, but still acknowledge that there is a virtue to the technical, philosophical usage that we need to focus on here. Without a clear semantic definition, it becomes difficult to evaluate the appropriate methodological stance we need to take concerning its measurement. Moreover, the standard philosophical notion makes it clear why measuring wellbeing is so important. We all have reason to care about our lives going well. This normative significance has guided the philosophical debate on the constituents of wellbeing, and plausibly should also guide our interest in measuring it. Secondly, as Hersch (2022) argues, we sometimes do compare across these different contexts, and such comparisons are meaningful. We can ask what is worse for someone: having a poor health outcome, or losing one’s financial stability. This indicates that there is a common meaning to doing well in different contexts that the concept of wellbeing simpliciter captures (see also Fumagalli 2022, 533-34). ⁶

Even if we agree on the semantics of wellbeing, there is a further fundamental issue that has been debated in the philosophy of wellbeing science. A common and widespread assumption within the philosophical literature is that there is a single theory of wellbeing that describes what constitutes wellbeing for all individuals. Alexandrova (2012a) suggests that this may be false: what constitutes wellbeing in a development contexts may simply be something substantively different from wellbeing in an industrialized country’s health context (cf. Lin 2017). ⁷ She calls this view Variantism (and its single-theory alternative Invariantism). This debate is more tangential to the debate between coherentism and foundationalism. ⁸ For simplicity, I will assume an invariantist account, but the arguments do not depend on this.

2.2. Theories of Wellbeing

Philosophers have developed theories about the fundamental constituents of wellbeing—what makes life non-derivatively good for those living it. Lange and Grünbaum (2023) call these characterizing theories of wellbeing. They take the following form:

WB theory: wellbeing is constituted by wellbeing good(s) X, Y, Z ….

I will refer to these goods as wellbeing goods, or constituents of wellbeing, and common candidates are pleasure, happiness, desire-satisfaction, achievement, friendship, and knowledge. To say that a good constitutes wellbeing is to say that it is necessarily true that when this good increases or decreases wellbeing increases of decreases as well.

Some theories may suggest there is only one such good. Hedonism maintains that only pleasure is a welfare good (or, more precisely, hedonic levels; e.g., Crisp 2006; Bramble 2016), and desire-satisfaction theories maintain that only desire-satisfaction is good (e.g., Bruckner 2010; Lukas 2009). ⁹ Finally, objective list theories identify wellbeing with a number of different goods, such as happiness, knowledge, and achievement (e.g., Fletcher 2013, 2016; Rice 2013).

2.3. Measures of Wellbeing

Measurement is a process of quantification with the aim of arriving at knowledge claims about a specific target concept (Chang 2004; Tal 2013). In the current context, however, we are particularly concerned with the empirical process of measurement: arriving at quantitative epistemic knowledge claims about the world, based on measurement instruments. Such knowledge claims only make sense if the target concept has some quantitative structure. Measurement theory typically distinguishes different levels of measurement: nominal, ordinal, interval, and ratio. Because it is generally assumed that someone’s wellbeing can be higher or lower than the wellbeing of that person at different times, but also higher or lower than the wellbeing of others, wellbeing is generally taken to have at least an ordinal structure that is intra- and interpersonally comparable. ¹⁰

Social scientists have, partially inspired by philosophical theories, developed wellbeing measures. Using a different measure of wellbeing will lead to different results (see Fumagalli 2013, 2019). Subjective wellbeing measures, for instance, ask individuals to rank their happiness on a scale—e.g., from 1 to 10, or a Likert scale. They are typically distinguished between measures that ask individuals to rank their happiness or satisfaction about their life as a whole (global measures) and measures that ask individuals to rank their experience at particular moments (momentary measures). While both are called subjective wellbeing measures, these different measures can lead to different qualitative results (e.g., Deaton and Stone 2013). Global measures are distinguished into life-satisfaction measures and happiness measures, which, in turn, sometimes result in different outcomes (e.g., Hansen 2012).

There are also preference-satisfaction measures (Benjamin et al. 2014), or value-based measures (Al-Janabi et al. 2012; Anand et al. 2009), which ask individuals to rank aspects, capabilities, or options, related to their life, which are then used as weights to aggregate these different aspects into a wellbeing measure. Such measures and subjective well-being measures result in different outcomes (Benjamin et al. 2012). Finally, there are also measures of wellbeing based on the self-assessment of capabilities, which in turn lead to qualitative results that can differ from other measures of wellbeing (Margolis et al. 2021; Van Ootegem and Verhofstadt 2012, 2016).

Some of these measures appear to map directly onto philosophical theories. Measures of subjective wellbeing seem to fit perfectly with hedonist conceptions of wellbeing, while preference-based measures seem to fit perfectly with desire-satisfaction, or preference-satisfaction theories of wellbeing. It would make sense to think that their efficacy as wellbeing measures depends purely on the plausibility of these underlying theories. However, it is perfectly possible that the best philosophical theory of wellbeing may not directly provide us with a recipe for the best measure of wellbeing. Daniel Hausman (2011), for instance, argues that preference-measures of wellbeing may provide evidence about wellbeing, even if preference-satisfaction theories of wellbeing are false, and Wren-Lewis (2014) develops a similar claim for subjective wellbeing measures. Because philosophical concepts may be difficult to operationalize, and because of the risk of measurement error, the best measure of happiness may be a preference-satisfaction measure, and vice versa.

Social scientists themselves also follow this type of reasoning. Pelin Kesiber and Ed Diener, for instance, acknowledge that wellbeing and subjective wellbeing may not be the same thing, but write “[i]t is reasonable to use subjective well-being as a proxy for well-being, even if it is not a perfect match” (Kesebir and Diener 2008 [2009], 62). In defense of measures based on happy life years, Ruut Veenhoven writes: “Note that I do not proclaim long and happy living as the essence of wellbeing, what I claim is that it is the most comprehensive indicator of this multi-facetted concept” (Veenhoven 2007, 226).

2.4. Foundationalism and Coherentism

We can now move on to our question of interest: How can we evaluate which measure is truly a measure of wellbeing, or which measure captures it best?

Coherentism and foundationalism are views on the justification of theories of the constituents of wellbeing that have important implications for how we can, and should, construct and validate measures of wellbeing. ¹¹ We can formulate both theories more precisely as follows:

Foundationalism: the reasons to believe in a particular view of the constituents of wellbeing are based on judgments on the constituents of wellbeing (e.g., intuitions and principles), but not on the basis of empirical facts about the correlates and causes of wellbeing measurements. The validity of wellbeing measures is justified with respect to their correspondence to what we have reason to believe constitutes wellbeing on the basis of our judgments.

Coherentism: the reasons we have to believe in a particular view of the constituents of wellbeing are based on judgments on the constituents of wellbeing (e.g., intuitions and principles) as well as descriptive facts about the causes and correlates of wellbeing measurements. A coherence between theoretical considerations and empirical findings provides us with reason to believe we are measuring what wellbeing is.

On the foundationalist picture, the justification of measures of wellbeing is necessarily a two-step process: we first determine what we have reason to believe constitutes wellbeing, and then formulate whether (existing) measures correspond to the wellbeing goods we have reason to believe constitute wellbeing. While these two steps need not follow each other chronologically, the justification only goes from theories to measures. For instance, if hedonism is true, and happiness is the only wellbeing good, we should assess whether measures of wellbeing plausibly capture, or can capture, happiness. The commitment to this two-step process follows from foundationalism’s commitment to the idea that our beliefs about wellbeing can only be justified by our judgments about wellbeing.

Critics have raised a number of challenges to this foundationalist picture (Alexandrova 2012b; Bishop 2015; Hersch 2022). For one, there is no agreement among philosophers about what the correct theory of wellbeing is. As Hersch writes: “One problem with adopting a foundationalist position in the context of well-being science is that in the current philosophical climate there is too much disagreement as to which well-being theory ought to serve as such a foundation” (Hersch 2022, 1051). Second, almost all philosophical conceptions of wellbeing are highly abstract. This abstraction often makes it hard to evaluate concrete measures. As Hersch writes: “We have theories of well-being and measures of well-being, but in the context of well-being science we must somehow coordinate between the two if we are to settle the question of ‘what is well-being?’, on the one hand, and the question of ‘how do we measure well-being?’, on the other” (2022, 1048). Measurement constructs typically deviate significantly from conceptions of wellbeing, even if the constructs are inspired by such philosophical conceptions. Third, according to foundationalism, the ultimate justification for judgments about the constituents of wellbeing are philosophical judgments. These may be based on pre-theoretical intuitions about wellbeing, as well as more theoretically informed ideas about what we judge the constituents of wellbeing to be (Bishop 2015). In other words, foundationalism ultimately justifies wellbeing judgments through standard philosophical methodology. As Bishop notes, however, this approach “assumes that our pre-scientific convictions are so worthy of belief that it is appropriate that they form the primary evidential base for a theory about the nature of well-being” (Bishop 2015, 26). This assumption, Bishop argues, may be too optimistic.

On coherentism, the justification of a wellbeing measure should be an iterative process. Not only do our judgments about the constituents of wellbeing provide a justification for measures of wellbeing, but empirical findings about these measures can themselves justify changing our theories about the constituents of wellbeing. As we will discuss in more detail below, the coherentist picture builds heavily on the success of such procedures in the history of natural science, where empirical findings have successfully led to changes in our understanding of the concepts scientists were interested in, such as temperature. As Hersch writes: “We need a theory of well-being to justify and be confident our well-being measure is right, and we need a well-being measure to verify and make use of our theory of well-being” (2022, 1059).

To clarify the crucial tension: both coherentism and foundationalism maintain that philosophical judgments about what wellbeing is should inform measures of wellbeing. The crucial difference between coherentism and foundationalism is that foundationalism does not believe that empirical findings on the causes and correlates of wellbeing have any implication for the question of what wellbeing is. Bishop puts this well. Foundationalism is committed to the following principle, while coherentism denies it:

The primary evidence for or against a theory about the nature of well-being comes from our judgments about the nature of well-being. Empirical findings about the causes, effects, or correlates of well-being are not relevant evidence for or against a theory about the nature of well-being. (Bishop 2015, 24)

In the following section, I will, contra Bishop, argue in favor of this principle in the context of wellbeing measurement. ¹²

3. Circularity and the Rationale for Coherentism

There seems to be an obvious issue with coherentism: it seems problematically circular. As Hersch’s quote above makes clear, coherentism maintains that theories justify measures, and measures justify theories. Bishop similarly claims that “the causes, effects, or correlates of well-being” should be treated as input in theorizing about the constituents of wellbeing. But these causes, effects, and correlates are based on measures of wellbeing, and in order to trust that they are correct, we need to trust that these measures are reliable. But, that depends on what we take wellbeing to be. Take Bishop’s example that wellbeing measures correlate with commuting times. To say that commuting times thereby instantiate a lower wellbeing requires not just a judgment about facts—e.g., commuting times lower life-satisfaction (or preference-satisfaction)—but also an evaluative judgment—life-satisfaction (or preference-satisfaction) constitutes, or tracks, wellbeing. If—following this finding—we were to take as a desideratum that any theory of wellbeing should capture the idea that whatever wellbeing is, it should correlate with income, this would be a non-sequitur:

P1) life-satisfaction correlates negatively with commuting times.

C) Wellbeing correlates negatively with commuting times.

The missing premise is:

Pm) life-satisfaction is wellbeing, or, corresponds reliably to wellbeing.

But this premise cannot be assumed, because this is exactly what is at stake. It therefore seems that assuming it would be problematically circular.

Defenders of coherentism justify their position precisely by citing that successful measurement is always circular to some extent. How do we know a measure of temperature, for instance, is indeed a measure of temperature, instead of something else? Or, how do we know a measure of economic growth is really a measure of economic growth? The only way to know this, it seems, is to use another measure of temperature or economic growth, and check if the first and second measure align. However, if we had such a measure, we would no longer need an independent measure after all. Hasok Chang (2004) calls this the problem of nomic measurement:

(1) We want to measure quantity X.

(2) Quantity X is not directly observable, so we infer it from another quantity Y, which is directly observable. […]

(3) For this inference we need a law that expresses X as a function of Y, as follows: X = f(Y).

(4) The form of this function f cannot be discovered or tested empirically, because that would involve knowing the values of both Y and X, and X is the unknown variable that we are trying to measure. (Chang 2004, 59)

Fill in for X: temperature (or wellbeing), and fill in for Y: thermometer expansion (subjective wellbeing), and the problem is clear. The problem of nomic measurement appears fundamental. Nevertheless, scientists do have reason to believe that their measures of temperature and economic growth actually do capture the targeted concepts. Coherentism draws much support from successful iterative justification in such examples. It is therefore worthwhile to see how this iterative justificatory process works in scientific measurement generally, to see how it should work in the context of wellbeing.

The iterative justificatory process builds on descriptive implications of theoretical knowledge about the concept at hand: if temperature increases, we should feel warmer (and start to sweat), if economic growth increases, unemployment should decrease. ¹³ If our measures do not agree with these implications, we must either revise our measures (of temperature, economic growth, or unemployment) or revise our theories. As we adjust both our theories and measures, we achieve coherence between our theoretical claims and measurement procedures (Alexandrova and Haybron 2016). It is important to note that this process requires theories from which we can derive descriptive implications about the behavior of measurements: temperature measurements should correspond to sweating of individuals who are nearby, or measures of economic growth should correspond to employment.

Through this iterative process, we may learn simultaneously about the reliability of our theoretical assertions about concepts, and the way we measure these concepts. Chang favorably cites Foley to describe the core of coherentism: “Coherentists […] propose […] that beliefs are justified in so far as they belong to a system of beliefs that are mutually supportive” (Foley 1998, 157; quoted in Chang 2004, 223). For instance, through this measurement procedure, we have revised our concept of temperature, and learned that temperature is kinetic energy. It is important to note that through this procedure, empirical findings have altered our conception of temperature. Some circularity is thus involved, but this does not undermine the justification: iteratively, measures justify theories, and theories justify measures.

A coherentist reasons that if this is how temperature measurement is justified through this circular, but reliable process, then a similar iterative process may justify what wellbeing is, and how it should be measured. As Hersch puts this: “Chang explains how the supposedly problematic circularity of a coherentist approach to scientific progress is anything but that” (Hersch 2022, 1058). Circularity in the justification of measures and conceptions is therefore not necessarily problematic. However, as Chang writes, and Hersch acknowledges: “Of course, there is no guarantee that the method of epistemic iteration will always succeed […] Whether that is possible is a contingent empirical question for each case” (Chang 2004, 226-27; quoted in Hersch 2022). And, in fact, as I will argue now, for wellbeing, it cannot work.

4. The Naturalistic Fallacy Argument Against Coherentism About Wellbeing

4.1. Evaluative and Explanatory Theories

In order to see why the coherentist iterative method cannot work in the context of wellbeing measurement, it is important to make a familiar distinction between different types of knowledge claims. While all measurement procedures are intended to arrive at knowledge claims about a target, there may be two different reasons for doing this. One chief aim of measurement in science is to explain descriptive phenomena. Temperature measurements can help explain descriptive facts, such as why water starts boiling at 100° Celsius in many circumstances, while air pressure measurements help explain why not all water boils at 100° Celsius.

However, there are also other aims of measurements. Wellbeing is a normative, or evaluative concept, and wellbeing measurement in scientific practice has as its chief aim the evaluation of states of affairs. When social scientists investigate whether more income leads to more happiness, they are evaluating the value of income to individuals. Consider for instance Richard Easterlin’s seminal article on the question whether increasing in income will “improve the human lot?” (Easterlin 1974).

While there may be exceptions (e.g., Minkov et al. 2019), the chief aim of wellbeing measurement is not explanatory, but evaluative: to assess whether a particular state of affairs (having a certain income level, being a parent, etc.) is good for groups of individuals. The behavior of wellbeing measures may itself be an explanandum, but typically not an explanans. This makes a crucial difference for the promise of the coherentist iterative process.

The difference in the aims of measurement corresponds to different types of theories from which the concepts emerge. Theories of wellbeing are importantly different from scientific theories, because scientific theories are typically descriptive, predictive, or explanatory (or a combination of those). The primary aim of wellbeing theories is to justify evaluative claims, such as “a higher income improves people’s lives.” However, from the claim that “knowledge is good” nothing follows about human behavior toward knowledge: it does not imply individuals will seek knowledge, nor does it imply that individuals will hold any particular attitudes toward knowledge. This is different from scientific theories, which do contain claims about how subjects or objects causally interact in the world. As a result, a failure to explain or predict some descriptive fact is not in itself a decisive reason to reject a measure of wellbeing. After all, whether a measure of wellbeing can help us predict or explain phenomena tells us little about whether it correctly informs us on whether lives are improving or worsening.

To see the distinction between these roles that concepts may play, consider Bentham’s two distinct forms of hedonism (Bentham 1789 [1988]), two theories in which the concept of pleasure plays different roles. Ethical hedonism maintains that states of affairs are morally desirable to the extent that they are pleasurable, while psychological hedonism maintains that people act with the intention to maximize their pleasure. The latter claim, if correct, may help explain certain phenomena, such as why so many people buy ice cream (knowing it is not good for their health). If psychological hedonism were true, the measurement of pleasure may serve the purpose of explaining phenomena. However, if only ethical hedonism were true, pleasure measurements could not help us explain phenomena, because the content of our knowledge claim—the quantity of pleasure determines the moral desirability of a state of affairs—is evaluative and not descriptive. The claim “pleasure is good” for individuals cannot explain why individuals behave in any particular way. Psychological hedonism may explain why people behave a certain way, because it tells us something about people’s motivations, but ethical hedonism does not. Moreover, ethical and psychological hedonism are independent: the truth of the one does not imply anything about the truth of the other.

The difference between measurement with explanatory and evaluative aims leads to different criteria of success for particular measurements. If the aim of a measure is primarily to explain, a failure to explain counts against a measure. However, if the chief aim is to evaluate, a failure to explain phenomena does not count against a measure. If measurements of pleasure do not actually explain human behavior, this would not be an argument against ethical hedonism (see Williams 2018 for an overview of recent empirical work on psychological hedonism, including a discussion on how difficult it is to test). ¹⁴

4.2. The Naturalistic Fallacy Argument

We can now formulate my key argument for foundationalism vis-à-vis coherentism. The discussed coherentist iteration solution to the problem of nomic measurement requires theoretical assertions about measurement constructs that describe the behavior of the construct. If those exist, then, if we expect that a measure of construct X may explain phenomenon Y, because we have a theory that implies that it does, but construct X in no way correlates with phenomenon Y, either our theory or our measure must be incorrect. As we have seen above, such revisions lead to a coherence between theories and measures, and thereby achieve scientific progress either through improving theories, or improving measures.

However, normative, or evaluative, theories, such as theories of wellbeing, do not have any direct implications about the behavior of measurement constructs. As I argued above, the truth about the fundamental constituents of wellbeing, unlike the concepts of temperature and economic growth, has no descriptive implications about the behavior of wellbeing goods, such as pleasure. Mere descriptive claims (X correlates with Y) cannot be inconsistent with mere claims about goodness (X is good). Consequently, there is no sense in which such theories may fail to correspond to our empirical findings, and consequently, there is never a reason to revise either a wellbeing theory or measure as a result of some empirical finding. This step, however, is a crucial step in the coherentist iterative process. Without this step, the iterative procedure that justifies our temperature measurements cannot provide us with any reason to revise our concepts. Thus, the coherentist justification that justifies temperature measurement cannot similarly justify wellbeing measurements.

Take any claim about the causes or correlates of a wellbeing measure, such as the claim life-satisfaction correlates with commuting times. Now take any philosophical theory of wellbeing: hedonism, desire-satisfactionism, objective list theories, or hybrids. The claim about causes or correlates of wellbeing cannot provide us with any reason to revise any of these theories if we hold them, because it is an empirical claim, while theories of wellbeing merely make claims about goodness—about what constitutes a good life for someone. Because such findings cannot provide us with reasons to revise the theories, they cannot start an iterative process of revision, and coherence can only be achieved by adjusting the measures to the theories. Any evaluative implication derived from a descriptive claim about the causes and correlates of wellbeing measures remains problematically circular.

This idea is not at all novel. It is simply a version of the naturalistic fallacy: from an “is” we cannot derive an evaluative claim—a claim about what is good—directly. Key here is that theories of wellbeing are merely claims about a particular type of goodness (X is good, in a wellbeing sense), while causes and correlates of measures are descriptive claims (claims about what “is”). For a claim about what “is” to have implications for what “is good,” we need at least one premise about what is good. In other words, any empirical claim about causes and correlates of a wellbeing measure can only have implications about wellbeing if we already start with an assumption about what wellbeing is. By themselves, they have no implication about wellbeing. Vice versa, an “is”—a descriptive or explanatory claim about the world—can also not be derived from a claim about goodness, without a premise that links this claim about goodness to something descriptive. Descriptive claims about the behavior of measures—e.g., life satisfaction correlates with commuting times—are independent of evaluative claims about the constituents of wellbeing.

The only way that empirical findings can have implications for what wellbeing consists in is to assume that they do bear some relation to wellbeing, but that is exactly what is at stake (Ingelström and van der Deijl 2021). Take the claim that wellbeing consists (partly) in X, in virtue of the fact that a scientific construct of X correlates with commuting times, without having independent reason to believe that wellbeing correlates with commuting times. ¹⁵ This begs the question.

4.3. Objection: We Have Both Evaluative and Descriptive Theories of Wellbeing

At this point, someone may object by pointing out that while philosophical theories of wellbeing are purely evaluative, we also have some descriptive assertions (if not theories) about wellbeing. We, for one, know that wellbeing is low for those who are suffering deep emotional pain, such as the loss of a loved one, and high for those who are spending enjoyable time with their loved ones. Moreover, even if we do not have full-fletched descriptive theories of wellbeing now, we may very well develop them in the future.

P1. People are doing better when they are wealthier.

P2. Wellbeing measure x correlates highly to wealth.

P3. Wellbeing measure y correlates poorly to wealth.

First conclusion: measure x is a superior measure of wellbeing than measure y.

Second conclusion: the notion of wellbeing underlying measure x is a better conception of wellbeing than the notion of wellbeing underlying measure y.

This line of reasoning comes close to the psychometric approach to the validation of psychological constructs (Alexandrova and Haybron 2016; Cronbach and Meehl 1955). It seems a plausible line of reasoning, and it seems to show that empirical findings should sometimes impact our belief in conceptions about the constituents of wellbeing. However, it is important to note that this line of reasoning depends entirely on our confidence in P1, which serves as a bridge between the evaluative and descriptive claims. The correlational findings in P2 and P3 cannot by themselves have any implications without the first premise.

We should note, however, that premise 1 packs together ideas about what wellbeing is, at least roughly, together with causal claims. We can separate P1 in two judgments: P1a states that wellbeing consists roughly of goods X, Y, Z, …; P1b states that X, Y, Z, … are served by wealth. P1a is now a purely evaluative claim. If only P1a were true, while P1b were false, neither of the conclusions would follow. This set of claims would have no implications for which accounts of the fundamental constituents of wellbeing is correct. P1b, in turn, is a causal claim. Importantly, however, P1a and P1b are epistemically independent. The truth of P1a implies nothing about the truth of P1b and vice versa. Seeing this, we can now once more see that it would be circular to derive any implications for the debate about the fundamental constituents of wellbeing from the observations in P2 and P3. A measure x may correlate better with wealth than measure y, and this may have implications for the truth of P1b. However, whether this has any implications for accounts of the constituents of wellbeing depends solely on P1a, a judgment about the constituents of wellbeing. Whether P1a is true, however, is precisely what is ultimately at stake. In other words, whether P2 and P3 have implications for the constituents of wellbeing depends on the constituents of wellbeing.

4.4. The Naturalistic Fallacy Argument in Sum

The naturalistic fallacy argument thus shows that empirical findings cannot themselves have any bearing on the truth about what wellbeing is. They only have implications about what wellbeing is if we first make assertions about how well a particular measure captures wellbeing. Justifying a theory of wellbeing on the basis of empirical findings about the behavior (i.e., correlates and causes) of wellbeing measures will therefore always be question begging. The fact that circular reasoning can provide us with good justificatory reasons in the context of some other measures, such as temperature measures, has no implications for the measurement of wellbeing and other value concepts, because the truth about what wellbeing is simply has no descriptive implications about the behavior of a measure of wellbeing.

5. Arguments Against Foundationalism

Even if coherentism faces the naturalistic fallacy argument, does foundationalism not face issues of its own? A defense of foundationalism in this context cannot merely present an argument against coherentism, it should also reject the arguments that have been voiced against foundationalism. In this section, I address and rebut the three arguments that have been levelled against foundationalism about wellbeing measurement that I discussed in section 2.

5.1. Lack of Agreement

A challenge that Hersch (2022), Bishop (2015), and Alexandrova (2012b) all pose against foundationalism is that there is no agreement between philosophers. It is important to note that at least some critics, such as Alexandrova (2012b), explicitly claim that this need not be a detrimental failure of the approach. However, all agree that it is a pragmatic challenge to the approach.

This objection, insofar as it is an objection, fails. First, a lack of agreement among philosophers is no objection to foundationalism itself. Foundationalism maintains that the validity of wellbeing measures can only be assessed on the basis of fundamental value judgments, which philosophical theories formalize. If there is no agreement on these value judgments and theories, there will be no agreement on the validity of these measures. Is that not a significant problem for foundationalism? No. In fact, this seems like the appropriate response. It would be problematic if the question of what the constituents of wellbeing are would be a highly controversial issue, but there was agreement on how to measure it. Wellbeing is a controversial evaluative concept, and the measurement of controversial evaluative concepts is bound to be controversial itself.

So, how should a science move forward when it aims to measure a controversial evaluative concept? There may be numerous ways to do so that are compatible with a foundationalist approach to wellbeing measurement. One possibility is that different scientists may simply adopt what they take to be the most plausible theory of wellbeing and operationalize it as best possible. For instance, if Wayne Sumner’s Authentic Happiness conception of wellbeing is taken by some to be the most plausible conception of wellbeing, existing life-satisfaction measures may be further refined to fit even better with this conception. While the justification of such measures will depend on the plausibility of this theory, many philosophical theories of wellbeing are plausible to some extent, even if there is no consensus. The successful operationalization of a plausible philosophical theory of wellbeing would give the measure at least some plausibility as well. Moreover, wellbeing research can adopt a pluralism about measures of wellbeing. Particularly when findings, perhaps qualitatively, hold across different measures, we can safely conclude that these findings are likely correct. Another approach is to use commonalities between wellbeing theories. Different wellbeing theories can agree on some constraints, or some guidelines, in formulating measures of wellbeing (e.g., van der Deijl 2017).

Controversy about what wellbeing is may be a challenge to wellbeing measurement, but it is a mistake to think that on foundationalist accounts of wellbeing measurement, no progress is possible until agreement is reached (see Fumagalli 2021).

5.2. Foundationalism Leaves an Unbreachable Gap Between Theory and Measurement

Generally, philosophical theories of wellbeing, in particular ones that are able to account for everyday intuitions about wellbeing, are abstract and difficult to operationalize. Measurements of wellbeing are not typically directed at wellbeing goods that popular philosophical theories of wellbeing posit—such as Sumner’s (1996) authentic happiness or informed-rational desire-satisfaction—but rather, at intermediate concepts such as subjective wellbeing and preference-satisfaction. The reason for this is that many abstract concepts are much more difficult to operationalize. For instance, while actual preferences are relatively easy to measure—by asking people, or observing their choices—measuring informed or ideal preferences is much less straightforward (Fumagalli 2022, 532). How could we learn what a particular subject would want if they were rational and informed? This presents the following problem for foundationalism: philosophical theories are too abstract to be operationalized, so, even if social scientists would use philosophical theories as their guiding principles, they would not be able to measure wellbeing on their basis. Hersch calls this the coordination problem. As Hersch writes:

Foundationalism does not acknowledge that coordinating between theory and measurement poses any difficulties, nor does it acknowledge that overcoming the problem and coordinating between theory and measurement requires working from both directions simultaneously. (Hersch 2022, 1051)

This argument misses its mark. A foundationalist can acknowledge that it is difficult to bridge the gap between abstract philosophical theories and measuremental practice. Consequently, it must acknowledge that measuring wellbeing may be difficult, and no measure of wellbeing may be perfectly able to capture wellbeing in all contexts.

It is not clear, however, why this would be a reason to reject foundationalism. We cannot reject a methodological view of how to measure wellbeing because it implies that there may always be a gap between measures of wellbeing and wellbeing itself. Rejecting the view on this basis would be tantamount to wishful thinking: of course, it would be great if wellbeing were perfectly measurable in a wide variety of contexts, but if we have reason to adopt a methodological approach that implies that this is not possible, we cannot reject the approach on this basis. ¹⁶

5.3. Foundationalism Puts Too Much Faith in Philosophical Methodology and Common Sense Intuitions

This brings us to our final criticism. According to foundationalism, what constitutes wellbeing cannot be learned through scientifically measuring wellbeing, and can therefore only be learned through philosophical analysis. ¹⁷ While philosophical methodology may itself be controversial, by and large, there is a widespread agreement that, at least in the case of wellbeing, and other value concepts, philosophers use a broadly construed reflective equilibrium approach. As Tiberius describes it:

(1) we start with a theory that purports to make sense of all the relevant considerations (the various intuitions, principles and background theories, i.e., the data);

(2) considerations that conflict with this theory are presented as objections to the theory; and

(3) we modify the theory to meet the objections, explain why the objections needn’t be heeded in the first place, or reject the theory entirely and start over. This process is repeated until we have answered all the objections and any further modification to the theory would result in conflict with other, more weighty considerations. (Tiberius 2013, 320)

A final criticism of the approach is that foundationalism places too much faith on the traditional approach. It is clear that this methodology relies heavily on the intuitions that individuals have. This is exactly what Bishop finds problematic:

The main problem with the traditional approach is that it begins with an exceptionally high degree of faith in the quality of our commonsense judgments about well-being. (Bishop 2015, 20)

Intuitions have a particular epistemic role in the standard methodology: They play a role in the justification of philosophical theories, and are the bedrock of philosophical reasoning. They are also difficult to challenge: if two individuals have a different intuition about the claim that a particular pain sensation is non-derivatively bad for someone, there is little that can rationally compel the one to agree with the other (Chalmers 2014). The badness of this pain state will be obvious to one person, and obviously false to the other. We also know that these intuitions are not universally shared and not very stable within individuals, and this leaves space for multiple theories of wellbeing, which are all inconsistent with some set of intuitions that some people have. There is no prospect of resolving these inconsistencies on the foundationalist methodology. This seems problematic.

There are two responses. A first is that the situation may not be as dire as it seems. As Fumagalli writes: critics of foundationalism “seem to significantly overestimate both the degree of disagreement between the best available philosophical theories of well-being and the degree of agreement between such theories that is required to get the science of well-being ‘off the ground’” (Fumagalli 2022, 528). A second is that there simply are no viable alternatives. Let us discuss these in turn.

Philosophers tend to focus on counterexamples to generate intuitions that count as arguments against their theories. This typically follows the following reasoning pattern:

(1) If wellbeing is constituted (only) by good X, good X should in all instances constitute wellbeing.

(2) In instance i, wellbeing is not constituted by good X (e.g., because a person who has this good is still doing badly).

(3) Therefore, wellbeing is not constituted (only) by good X.

Step 2 is typically based on a counterexample that triggers an intuition. Probably the most well-known counterexample in the context of philosophical debates about wellbeing is Nozick’s experience machine example. Nozick imagines a machine in which one can experience all sorts of pleasurable experiences but takes it as evident that such a life is not that great, and consequently hedonism (according to which only pleasure makes life good) is false. Now, it turns out that it at least seems that people have widely varying intuitions about whether life in a simulated reality can be good for the one living it (De Brigard 2010; Hindriks and Douven 2018; Weijers 2014). This may induce pessimism about the methodology: if people have different intuitions, which intuitions should matter for wellbeing?

However, what the focus on counterexamples leaves out is that there is in fact a wide agreement on wellbeing judgments among philosophers. For example, those in pain, depressed, and frustrated are faring badly, enjoyment is (at least typically) good for individuals, and money has no intrinsic value. There are no theories of wellbeing that intrinsically value the number of children one gets, or one’s social status. The fact that not all intuitions are shared obscures the fact that the significant majority of them are. In fact, if this were not the case, formulating counterexamples would be a pointless exercise, as no one would agree on the intuitions about these cases. However, exactly because many people do agree on these counterexamples pose a problem. A hedonist may not share the intuition that the experience machine does not provide the best possible life, but must understand the pull of the case to some extent, as even hedonists acknowledge it as a problem (e.g., Bramble 2016; Crisp 2006).

Second, even if intuitive judgments are unreliable, this is only an argument against foundationalism if there is a viable alternative. Intuitive judgments of the type standard philosophical methodology is based on cannot be avoided if we want to arrive at value judgments about wellbeing. How could we know what wellbeing is besides reasoning about it starting from basic judgments about it? Bishop provides the clearest worked out alternative:

The inclusive approach gives us two simple tests for knowing when we have found the correct theory of well-being: When philosophers build their various accounts of well-being, the true theory will imply that they are all successfully describing wellbeing, even if they have some of the details wrong. And when psychologists use their various methods to study well-being, the true theory will imply that they are all studying well-being, even if they have some of the details wrong. (Bishop 2015, 2)

The inclusive approach takes as a given that the empirical study of wellbeing is indeed about wellbeing. What kind of justification can we offer for this? We may have a premise that all scientists are at least roughly right about the concepts they study. The measures that scientists have of concepts, such as wellbeing, are always indeed about wellbeing. However, this would be highly unsatisfactory, and circular. The answer cannot be that they are, because they are measures of wellbeing, or that they are, because they were developed by scientists.

A more plausible position would be that we have some faith that scientists are correct, because they have thought well about their constructs, and their relation to wellbeing. For instance, it may just seem clear to scientists that if individuals themselves believe that they are doing well, they must be doing well, at least in a reasonably reliable way. In fact, this is exactly the type of reasoning that several prominent wellbeing researchers engage in.

In the introduction, I cited Ruut Veenhoven and Ed Diener, who argued that they find it independently plausible that when individuals believe they are doing well, they actually are doing well, regardless of whether subjective wellbeing actually constitutes wellbeing. These may be good reasons to take subjective wellbeing to be an indicator of wellbeing, but they are also reasons that stem from intuition, not empirical fact. It is a different kind of intuition than the type of intuitions that philosophers tend to work with: it is an intuition about what is a good indicator of wellbeing, rather than about what wellbeing is, but it is nevertheless an intuition about wellbeing.

Bishop considers the idea that we have reason to believe that constructs about wellbeing are only justified by intuitive judgments of scientists, but rejects it:

A critic might argue that the inclusive approach does not rest on firmer evidential foundations than the traditional approach, since it trades the potentially confused intuitions of philosophers for the potentially confused intuitions of philosophers, psychologists, and laypeople. But this is not an accurate description of the inclusive approach. The scientific evidence does not simply reflect a group of people’s commonsense judgments. We can’t recreate the scientific evidence by asking psychologists what they think about the nature of well-being. The inclusive approach assumes that the explosion of scientific research on well-being over the past two decades has taught us something about well-being, a real state of the world. (Bishop 2015, 34)

Bishop’s defense is puzzling. Bishop starts from the claim that the study of wellbeing is indeed about wellbeing. But this is exactly what is at stake when we ask to what extent scientists who study wellbeing capture wellbeing adequately. We may have good reasons to believe that the study of wellbeing is indeed about wellbeing, but those reasons rely on the type of claims that Diener and Veenhoven make: intuitive justifications about the plausible relationship between measures and wellbeing at large. Without such justifications, there is no justification for the claim that wellbeing research is indeed about wellbeing.

If we want to evaluate the plausibility of a wellbeing measure, there is no getting around the fact that at some point our justification must rely on bedrock evaluative judgments about goodness: intuitive judgments. Empirical facts do not, by themselves, entail anything about such intuitive judgments. It is incorrect to say that foundationalism puts too much faith in standard philosophical methodology based on common sense intuitions, because it can acknowledge that such judgments may in fact not be very reliable. What it is committed to is the claim that there simply is no viable alternative for the evaluation of wellbeing measures.

6. Conclusion

How can we evaluate which, if any, measure of wellbeing captures wellbeing well—how good life is for the individuals living those lives? Foundationalism states that we can only answer this question by looking at our judgments of what we take wellbeing to be, and then assess whether a measure fits well with our conception of wellbeing. It maintains that empirical facts about the causes and correlates of different wellbeing measures have no implications for what wellbeing is. I have argued that for evaluative concepts, like wellbeing, foundationalism must be correct, because empirical facts about wellbeing measures have no implications for the truth of any conception of wellbeing. I have rejected three arguments against foundationalism: (1) that foundationalism is false because philosophers do not agree on what wellbeing is, (2) that foundationalism leaves an unbreachable gap between wellbeing theories and wellbeing measurements, and (3) that it puts too much faith in standard philosophical methodology.

This argument has important implications for the science of wellbeing. Foundationalism may seem like a position that endorses disciplinary boundaries, but in fact, it implies that in order to study wellbeing philosophy and empirical sciences need to cooperate. It shows that deep philosophical questions underlie every empirical claim about wellbeing. At the same time, we should acknowledge that philosophical wellbeing theories, by themselves, are not very helpful in practice. We need empirical science to learn about the causes, effects, and distribution of wellbeing in society.

Foundationalism also shows why developing wellbeing measures is not straightforward, and why it is quite different from formulating philosophical theories of wellbeing. Not only do claims about wellbeing rely on claims about what wellbeing is, but they also rely on claims about how closely that which we can measure (e.g., measures of subjective wellbeing) relates to wellbeing, and in which contexts these measures are good indicators of what we take wellbeing to be. This exercise is something neither philosophers nor social scientists are trained in. To move forward, we need social scientists who can reason philosophically, and philosophers who understand empirical science.