Better-Making Properties and the Objectivity of Value Disagreement

2.1 Complex better-making properties

Multiple features complicate matters. Comparing the knives in (3b), a might turn out to be sharper and have a better handle than b, and therefore it may be more suitable than b as a kitchen knife. It is well-known that there are many ways to combine features in such a multidimensional scenario. If the features can be expressed quantitatively, one might sum them up, provided that intuitions about overall comparisons remain compensatory and consistent with additive models. For example, a knife with handle quality 3 and blade sharpness 5 must be equal in value to a knife with handle quality 5 and blade sharpness 3 in an additive model. Sometimes, such additive models do not suffice and more complicated aggregation methods are called for. We may put aside many of these details, however, because better-making property are allowed to be arbitrarily complex. Better-making properties are decisive for a comparison if all other relevant comparison features are equal, no matter how complex they are. Some such comparisons may be between hypothetical objects that only differ in one aspect.

Some cases deserve special attention, though. Several properties may be decisive only when they are present together in a sub-additive or a super-additive way. Super-additivity means, in this context, that if those features could be quantified, then their combined presence would have a higher value than the sum of the values of each of the features taken individually. Going back to Moore (1903), this view is often discussed under the label “organic unity”.5 In contrast, in a sub-additive value combination the combined presence of the features may have a lower value than the sum of the values of each feature taken individually. The holistic assumption behind sub- and super-additive value aggregation can be reformulated as the thesis that the complex better-making property emerges as a qualitatively new property. Claiming that such properties exist ought not pose more problems than the appeal to the holistic assumption.

Better-making properties cannot be contradictory under the same value. If there are two properties, P and P', and two items a and b such that P(a) & ¬ P(b) & ¬ P′(a) & P′(b), then P and P′ cannot be better-making properties belonging to the same value. This constraint is more of a methodological requirement than one concerning value philosophy. Methodologically, it makes sense to specify that conflicting better-making properties belong to separate (sub)values, because methods for aggregating multiple value relations into an overall assessment already allow for dealing with such value conflicts. Otherwise, the underlying value representations would have to be paraconsistent, allowing for the truth of a ≻ b & b ≻ a, rather than the unproblematic case a ⪰ b & b ⪰ a commonly used to define a ∼ b. Paraconsistent logics of value can represent moral dilemmas. Still, an account with multiple dimensions has enough expressive power without additional paraconsistency if it allows aggregation failures to represent incomparability. Assume the knife a is better than b in terms of sharpness and b is better than a in handle quality. Various value aggregation algorithms provide solutions to this problem. If the sharpness aspect weighs more than or outranks the handle quality, a might be better than b. If the two values have the same weight or rank, then a ∼ b would be an acceptable aggregation. Finally, it is feasible to have two values in a conflict so that aggregating them fails in a specific case.

There are additional technical requirements on the rules for better-making properties. They must generally cohere with the properties of the value relations they indirectly constitute when assembled from piece-wise comparisons. Strict betterness ≻ is often considered transitive.6 If this is the case, then the following rule must hold: For any three objects x, y, z, if there is a better-making property P1 that implies x ≻ y, and there is a better making property P2 that implies y ≻ z, then there is a better-making property P3 such that x ≻ z. Without further ado, the above rule also complies with the irreflexivity of strict betterness since P(a) & ¬ P(a) is already excluded as a contradiction when the base logic is not paraconsistent. The standard account of “better than” does not require other rules, but when using nonstandard value relations like semiorders, additional rules must ensure that better-making properties comply with those alternative base relations. For instance, semiorders have the “Ferrer’s property”.7

Finally, we should avoid trivial positions. A better-making property for value comparison a ≻ b may not be circular. We should not allow properties whose comprehensive characterization would amount to restating the value comparison in the subsequent of the rule. For instance, this condition prohibits the property of being better than b. Although a better-making property can be relational, it may not be relational in the trivial sense of repeating the same or a similar value relation that represents the value under discussion.

2.2 Better-making properties and final value

Better-making properties seem to be hard to square with intrinsic and final value. Since there is widespread agreement in the Moorean tradition of axiology that final value exists, this criticism would at least severely limit the usefulness of the above definition. The purpose of this section is to show that better-making properties are compatible with final value.

Something has a final value when it is valuable for its own sake, without having to take into account other values and consequences of having the value. For example, if friendship in (1) has final value, it is not valuable because having friends provides pleasure or other advantages, it is valuable for its own sake. Some philosophers, such as Korsgaard (1983), consider what is valuable for its own sake final value and oppose it to instrumental value, whereas intrinsic value is opposed to extrinsic value and based on intrinsic properties. This terminology makes final value more important than intrinsic value because there are compelling examples of things with final value not based on an intrinsic property (see Beardsley 1965; O’Neill 1992; Kagan 1998; Rabinowicz and Rønnow-Rasmussen 2000, 2005). Even authors like Zimmerman (2001), who prefers the label “intrinsic”, agree that intrinsic value cannot always be based on intrinsic properties of comparison objects in the narrow sense.

For example, according to Beardsley (1965) rare stamps may have a value on their own, and being rare is not an intrinsic property of a stamp. Zimmerman solves this problem by delineating an ontology of states of affairs with basic intrinsic value, but we need not enter the (mostly terminological) debate about intrinsic versus final value. It suffices for current purposes to acknowledge that among arbitrary comparison objects, not all final value is based on intrinsic properties.8 Likewise, it need not concern us that some authors like Zimmerman (2001) and Perrine (2018) argue that the basic objects of comparisons are states of affairs, whereas others such as Rabinowicz and Rønnow-Rasmussen (2000) argue against this view. The following discussion is neutral about the nature of the comparison objects.

The criticism is this: A better-making property provides the reason why one comparison object is better than another; that is a comparative definition. In contrast, final value does not seem to be comparative at all. To say that friendship in (1) has final value is to say that it is valuable on its own and not relative to other concepts. Hedonists consider pleasure a final value not because it is better than pain but because it is intrinsically good from their point of view. A painting might be valuable in its own right, being so unique that it would be hard even to compare it to other paintings. Such examples seem to indicate that better-making properties cannot provide a final value and, therefore, cannot be the sole reason why we attribute value in general if final value exists, although they may be useful for reasoning about the instrumental and extrinsic value of objects. As I will argue, this criticism rests on a misunderstanding. Any kind of value, including final value, must allow for comparisons, and better-making properties provide reasons for specific comparisons. There is no incompatibility in the first place.

My counter-argument relies on the choice-guiding nature of values. A necessary, though not sufficient condition for being a value is to potentially guide someone’s choices. That is to say, a particular value might never guide anyone’s choices in practice, but if someone has to choose between several alternatives, then the value must be able to guide the choice provided it is applicable and relevant. I consider this an analytic aspect of what it means to be a value. There are no values that cannot possibly be choice-guiding.9 The person in need of guidance must somehow be able to apply or use that value to evaluate alternatives and figure out, based on that value, whether one alternative is better than another, they are equally good under that value, they are on a par in the sense of Chang (2002), or the comparison fails for some reason. In all cases except the last one, the properties that provide intrinsic value to a comparison object must play an integral role in the comparison since they are the reasons why these objects have value relative to the other object, and these reasons should guide choices rather than something else.

Thus, when something has a final value, the properties that give it this value must allow for comparisons. When comparing, a better-making property may be identical to the property or relation that lends the comparison object its final value. Nevertheless, the fact that a comparison is made need not be constitutive of the value. For example, suppose that two states of affairs a and b containing John and Mary are compared. Suppose John and Mary are good friends in a and no friends in b. If friendship has intrinsic value, then one might say that a is better than b because a has the property of containing two friends that b lacks. This property is the better-making property in this example. Despite this, the fact that a and b are compared is not itself constitutive of the intrinsic value of friendship.

For Moore (1922, 260-261), intrinsic value can come to a specific degree, which trivially enables multiple comparisons. Zimmerman (2001, 159-180) expands on this and even argues that value can be summed up. These fairly strong assumptions about value allow one to use utility functions to represent value. I will address some problems with such representations in Section 3.2 when discussing desire. For now, it suffices to show that better-making properties remain compatible with such views on intrinsic value.

Suppose a in the above example has the intrinsic value of friendship to degree 0.8 on a normalized scale between 0 and 1, and b has this value to degree 0 because there is no friendship at all in this state of affairs. The better-making property is the property of containing friendship to a normalized degree 0.8 (whatever that means). The same better-making property would also serve as a reason for the comparison to a third state of affairs with two more superficial friends of degree 0.4 only, yielding the judgments a ≻ c ≻ b. The better-making properties include the particular degrees or amounts of the intrinsic value in such cases. Although it is doubtful that such an account of intrinsic value would be adequate for examples like (1) and (2), and one might argue instead that such examples only involve ordinal value comparisons, better-making properties are perfectly compatible with stronger value conceptions according to which intrinsic value comes at a degree.

In summary, better-making properties neither implicitly nor explicitly presume that comparisons are value-constitutive. Value must be able to guide someone’s choices under the right circumstances and allow for comparisons, yet the reason why something has value may still be that it has value for its own sake.

3.1 Lack of disagreement about matters of personal taste

This section aims to show that apparent disagreements about personal taste are not value disagreements since they are no disagreements. This idea is not new; it has been discussed quite extensively in recent literature on relativism versus contextualism of predicates of personal taste.

Consider a disagreement in the ice cream scenario. As Lasersohn (2005, 2008) argues, disputes involving uses of predicates of personal taste may be cases of faultless disagreement. Alice might truthfully state (4), and Bob might truthfully state the negation of this sentence. Both assertions may be true, respectively, in relation to the assessors Alice and Bob. According to Lasersohn, in such a case the disagreement is faultless; both of them are right. Other people may also assess the statements in one or the other way in this version of relativism.

It is controversial whether such statements are true relative to an assessor (assessor-relativism) or whether their truth-value varies only because their semantic content varies (contextualism).11 We do not have to decide on these issues, as both accounts share the same idea: If a comparison is based on preferences of personal taste, it is subjective because people’s tastes differ. What is questionable about these cases is whether these cases count as instances of disagreement.12 As long as Alice in (4) provides as a reason that this is her preference, there need not be any disagreement between Alice and Bob precisely because subjective justifications are deemed appropriate in matters of personal taste. Suppose Bob prefers vanilla over chocolate ice cream. In that case, his preference is compatible with Alice’s preference, and he can agree with Alice if he agrees that (4) is based on her preferences. Strictly speaking, it is incorrect to call such cases subjective disagreements because they are no disagreements in the first place.

This is not to say disagreements over such issues cannot occur at all. A dispute might concern whether someone has a particular preference. Although there is some first-person authority about preferences, this authority is not absolute. Bob may know Alice’s preferences better than her. People only sometimes know what they want and can be mistaken or confused about their preferences. Moreover, people may signal disagreement in a conversation, even when there is no disagreement about the underlying subjective aspect of an evaluation. A dispute might concern something else, such as presupposed content or social inferences drawn from the belief that someone has a specific taste. For example, Bob may disagree with Alice because he believes that people who prefer chocolate ice cream over vanilla ice cream are tasteless brutes. As ridiculous as this may sound about ice cream, disputes about musical preferences are often of this sort.13 It is common in the personal, social, and political realms to have disagreements about something other than the content of a particular utterance the disagreement seems to be about. In these indirect disputes, the utterance content only serves as fuel for other persistent disagreements in the background.

There may also be disagreement over whether the justifications can be subjective. For example, one person may believe that there are objective criteria for determining if one painting is better than another, yet another may be a subjectivist about art. People may also dispute what constitutes a better-making property and whether objects have the property in question. However, once we identify a disagreement as one about taste, we know it will involve primarily subjective justifications. In the other examples mentioned, the disagreement concerns something else, such as social norms and functions. Such additional disagreements may be legitimate, but they are not direct disagreements about the evaluative statement in question. They concern the better-making properties, or a standpoint or social issue hidden behind the evaluative statements seemingly under dispute.

Considering all this, I suggest distinguishing between more broadly conceived evaluative comparisons and value comparisons in the narrow sense. Value comparisons, in the narrow sense, are not based on subjective preferences, although the underlying value relations may look similar to these from a modeling perspective. Value statements are meant to be intersubjective or objective. In contrast, apparent taste disagreements concern evaluative comparisons that reveal subjective preferences, but they involve no disagreement; if there is disagreement, it is not directly about the evaluative statement.

3.2 Better-making properties are objective

In this section, I argue that better-making properties are objective. As previously stated, agent-relative and relational properties can be objective. But what does objective mean? Although this question may be hard to answer in general, the following distinctions suffice for the purpose of this article. A subjective property is one that an object can only have if one particular person has a belief or a similar non-factive, truth-upholding attitude about the object and if the property cannot be reduced to a property that does not entail that attitude.14 In contrast, characterizing an objective property does either not involve any reference to attitudes at all or it involves factive attitudes like knowledge.

A property may also be intersubjective. If a property P is such that having P presupposes that rational persons within a given community with common knowledge about the world can be expected to hold certain attitudes dispositionally, or upon sincere reflection, about objects that have the property, then P is intersubjective.

To exemplify these distinctions, consider monetary cost. Being believed by Bob to cost €50 is a subjective property. So is being believed by Alice to cost €12. In contrast, the property of costing €50 is an intersubjective property. Monetary systems hinge on people’s attitudes about money and its worth, the governing institutions, and markets. In the case of fiat money, those beliefs partially constitute the property of costing €50. Nevertheless, the property of costing €50 is not constituted by any particular person’s belief about the object, not even the seller’s, and therefore is not subjective. Finally, being known by Bob to cost €50 is an objective property because knowledge is factive; everything with this property also has the property of costing €50, which does not require a specific person to hold a belief about it.15

Suppose a better-making property P was subjective. According to the definition of a better-making property, P(a)&¬ P(b) implies the value comparison a ≻ b. Since a person needs to hold a non-factive attitude about an object for that object to have a subjective property, the rule states in this case that it is a sufficient condition for the truth of a value comparison that a particular person holds an attitude about the object. This position is absurd if the attitude in question is belief or another truth-upholding attitude. The mere fact that someone believes something about a and does not believe the same about b does not warrant that a is better than b; there must be some property in which a and b differ that allows for that conclusion regardless of what a particular person believes about them.

Consider the monetary value of two comparison items, for instance. Just because Alice believes that a is cheaper than b and therefore better in terms of cost does not warrant the conclusion that a is better than b in terms of cost; a is only better than b under this value when it is cheaper. Under normal circumstances, it is not enough for someone to believe that the comparison items have or lack a particular property; they must actually have the property or lack it. If Alice happens to find out that her belief was false and b is cheaper than a, she would not say that her values (or, in this case, subjective evaluation) have changed. She would rather say that she misjudged the value of a in terms of costs and concede, insofar as she acts rationally, that b was better than a in terms of costs in the first place.

Only matters of personal taste might be an exception to this rule. Maybe Alice’s belief that some ice cream tastes like chocolate is good enough for her evaluation, even if her senses are confused and the ice cream does not actually taste of chocolate. However, as I have argued above, such examples do not illustrate value comparisons because they do not give rise to disagreement. A subjectivist may call these subjective evaluations values, of course. However, this is merely a terminological choice; the point is that subjective evaluations based on personal preferences differ substantially from value comparisons that constitute what one might call real or “genuine” values because the latter give rise to disagreements, whereas the former do not.

Properties involving attitudes that are directly about comparison items fare better. Could the property of being desired by someone be a better-making property? Such an account might seem plausible for Humeans who consider desire a basis for choice. However, there are compelling arguments against the idea that the property of being desired by someone makes something better.

To begin with, being desired does not suffice. To conclude that a is better than b, the desire for a must be greater than the desire for b. So degrees or intensities of desire are needed. If these exist, then it is indeed possible to formulate a rule stating that whenever the amount of X’s desire for a is larger than the amount of X’s desire for b, then a is better than b for X.

However, such conceptions of “better than” as desire get the direction of justification wrong. We desire a more than b because it is better (for us, to stay within the agent-relative realm for the sake of argument). The converse is not valid. It is not generally true that whatever we desire more than something else is better (for us).16 The reason to reject desire as a basis for goodness is not potential psychological confusion, as is sometimes argued against subjectivists, but rather a temporal dimension of desire that goodness does not have. We desire something episodically, at a particular time, when the consequences of fulfilling that desire are not yet fully known. If the consequences turn out to be negative in the future, the person still had the desire in the past.

In contrast, suppose we say that something is better than something else for someone. If the consequences turn out to be negative, the initial betterness statement is retracted and considered false. It is not the case that the option for that person was good and is now no longer good; rather, it was bad from the start. This asymmetry in the temporal dimension of the two notions makes it impossible to use desire as a substitute for goodness.

Suppose, for the sake of argument, an account built on a Desire Satisfaction Principle despite these flaws. The resulting position would render value relations obsolete. Utility functions from objects to real numbers can represent an amount of desire that allows for “greater than” comparisons. Desiring a more than b means that the amount of desire for a is greater than the amount of desire for b, i.e., u( a) > u( b) holds. According to the theory of scale types introduced by psychologist Stevens (1946) and formally worked out in measurement theory (see, e.g., Roberts 1979; Krantz et al. 1971, 1989, 1990), talking about amounts in this way means that the utility function u(.) rests at least on an interval scale and more likely on a ratio scale.17 A corresponding value relation can be extracted from such a utility representation in a mechanical way by defining x ≻ y ⇔_Def. u( x) > u( y) and x ∼ y ⇔_Def. u( x) = u( y). This construction makes the value relation dispensable and requires assumptions much stronger than merely talking about “better than” comparisons within a value. Utility functions guarantee that all value comparisons are complete and transitive, provided that additional constraints are met in case there are uncountably many comparison objects. Utility functions also make all value comparisons compensatory, which is a dubious assumption. To cut a long story short, desire understood in this way is a stronger value representation than a mere value relation. It makes the latter redundant.18

Even if one is willing to defend such an account, the property of comparison object a of being desired to amount u( a) by person X cannot serve as a better-making property. The comparison u( a) > u( b), not the amount of desire for a itself, makes a better than b, and this comparison violates the circularity prohibition of Section 2.1. Finally, even under a desire-as-utility view, when we ask why a particular object a is better than b in a given evaluation situation, the reason cannot just be that it is more desirable. Rather, a is more desirable because it has some property that b lacks. Desire is not blind, something in the desired object needs to spark it.

3.3 The objectivity of value disagreement

To recapitulate, objective better-making properties are sufficient conditions for the truth of value comparisons. These properties are also necessary for justifying value comparisons, so every justification of a value comparison has an objective component. However, one point of the previous sections was that these justifications are typically more exhaustive. Part of a justification may also concern what constitutes a better-making property for a particular value and how different values enter an overall value assessment. Finally, a disagreement may also arise over the relevance of specific values. For example, someone might deny that comparisons of a product’s packaging design ought to enter its evaluation. In contrast, someone else might insist that it is an essential aspect of the purchasing experience. Because of these additional possibilities, one might doubt that broader aspects of a justification need to rest on narrow facts.

Moral intuitionists and particularists like Dancy (2004) have expressed one such doubt. According to Dancy, there cannot be an overarching systematic theory that justifies moral judgments and morally relevant value comparisons. Our moral practices are too context-dependent and have too many exceptions to allow for general theories. Instead, we must rely on moral intuitions in each evaluative context. These enter broader justifications of value statements.

It is worth noting, however, that moral intuitionism and particularism are compatible with the approach presented thus far. Sometimes a justification may appeal to intuitions, and it is also possible to have different justifications in different contexts. Nevertheless, it seems doubtful that intuitions alone can be decisive for particular value disagreements.

The problem is that intuitions are not generally a source of evidence. I follow Hintikka (1999) in this regard, though my own take is a bit less radical. In my point of view, intuitions may provide evidence in moral philosophy due to certain anthropological constants, but I agree with Hintikka that they are methodologically useless for resolving disagreements. Suppose most people share roughly the same intuitions about a value statement. That means the value statement is uncontroversial, and most people agree about it. In that case, there is no demand for a justification, and there will be widespread agreement over the better-making properties. Such cases may occur, but they are of little interest in the light of error-theoretic arguments like those of Mackie (1977). Many interesting value statements trigger persistent disagreements. So suppose there are conflicting intuitions instead. Then intuitions themselves cannot resolve a disagreement, although they might help to address it. There are essentially three ways to deal with such cases:

According to the thesis defended in Section 3.1, the first response means that the alleged value statement does not concern value but only subjective preference. There is no fundamental disagreement between people who seemingly disagree about such statements, or the disagreement is about something else. The second response is likewise possible. However, it is a long stretch to claim these are the only possibilities. At least some value comparisons can reasonably be expected to fall into the third category. So what about the third case?

Factual disagreements can be persistent, and their resolution may require detailed domain knowledge. Nobody would expect non-specialists to be able to determine whether a statement in physics is reasonably well-confirmed or false; physicists do that, and they need to study physics for years to acquire the skills to judge and advance physical theories. Similarly, problems of what constitutes better-making properties and how to combine different values into an overall assessment might hinge on moral facts. Scanlon (2014) and Parfit (2011) defend moral facts based on “domain pluralism”, the thesis that the truth of statements and the existence of corresponding facts are established differently by different domains of inquiry. Science is concerned with narrow facts, mathematical reasoning is concerned with mathematical facts and the existence of mathematical objects, moral reasoning is concerned with moral facts, and so forth. If this view is correct, moral and axiological facts might make the theories that support value statements true or at least more adequate than other theories. Some of these facts might not be narrow in the sense introduced in Section 1.

Domain pluralism is controversial. What would these non-narrow facts be, and how do we access them? Are they like mathematical truths? This article does not need to answer these questions and decide whether domain pluralism is acceptable. Whether moral facts exist is independent of justifying and ranking the overall merits of theories that support value comparisons. Error theory, moral relativism, naturalism, non-cognitivism, and moral realism have one thing in common: Theories are not compared according to their moral value. We compare them according to how close we believe they are to being true, and, in a more practical sense, according to theory virtues and merits exemplifying (broadly conceived) epistemic value. There is no reason to believe that axiological theories work substantially differently than theories in other fields from an epistemological point of view. A justification of a value statement rests on a supporting theory and corresponding beliefs, which may include metaethical and normative stances, and any theory is ultimately assessed on the basis of its overall merits. Epistemic values decide the outcome of such an evaluation. Which justifications and supporting theories are most likely true? Which justification has best explanatory adequacy? Which one integrates best with other value-related issues and metaethical theories? Which one is internally most coherent? Justifying a value comparison requires answering these questions, which cannot be answered by intuitions alone.

So, the answer to the question of how to deal with questions of the third kind is that, ultimately, the epistemic merits of supporting theories decide between competing justifications of value statements. It is a separate question whether those merits reliably track moral facts and in which way, and it seems likely that viable answers to these questions vary from value to value. Different types of values have different supporting theories with different overall merits, and we need to address each of them separately.

Is this the trivial position mentioned at the beginning of this article? Although it remains close to it, the new position is no longer trivial. First, better-making properties are not trivial, and whether a comparison object has better-making property or lacks it depends on narrow facts. This aspect of value comparisons is objective. Value comparisons between hypothetical comparison objects are equally objective. In this case, law-like statements from well-confirmed theories allow us to derive the relevant facts about the comparison objects. For instance, to have any validity, causal consequences of hypothetical courses of action that give rise to better-making properties are based on law-like statements about the world, and the theories supporting these statements are empirical. Second, ranking theories according to their overall merits is far from being trivial, as the vast body of literature on abduction and inference to the best explanation illustrates.19 The epistemic evaluation involved in inference to the best explanation does not involve moral value. Even when non-narrow facts are involved in this evaluation, epistemic values trump other types of value and ultimately guide our judgments about value statements. All aspects of value disagreement are objective in this sense.

This position remains compatible with the view that there is sometimes no acceptable justification for a particular type of value statement. Judging that there is no acceptable justification is itself an evaluative position, though one that might remain agnostic about the original value statement. In that case, the proper response acknowledges that there is no corresponding value. This response is similar to how we (epistemically should) deal with existence claims in other domains of inquiry. For example, as Russell (1952) famously pointed out in his rejection of theism, the claim that there is a teapot flying in orbit between Earth and Mars has no good enough justification, so the default assumption ought to be that there is no such teapot. Likewise, if there is no good enough justification for a value statement, the default assumption is that there is no underlying value.