Hostname: page-component-74656f9884-8r8nv Total loading time: 0 Render date: 2026-11-07T12:09:36.131Z Has data issue: false hasContentIssue false
  • English
  • Français

Value and Preference Relations: Are They Symmetric?

Published online by Cambridge University Press:  21 September 2015

MAURO ROSSI
MAURO ROSSI*
Affiliation:
Université du Québec à Montréal rossi.mauro@uqam.ca
Rights & Permissions [Opens in a new window]

Abstract

According to Wlodek Rabinowicz's fitting-attitude analysis of comparative value, it is possible to analyse both standard and non-standard value relations in terms of the standard preference relations and two levels of normativity. In a recent article, however, Johan Gustafsson has argued that Rabinowicz's analysis violates a principle of value–preference symmetry, according to which for any value relation, there is a corresponding preference relation. Gustafsson has proposed an alternative analysis which respects this principle and which allegedly accounts for the idea that originally motivated Rabinowicz's analysis, namely, that in some cases different preference relations between a pair of items are equally permissible. The goal of my article is to show that the arguments offered by Gustafsson in favour of his account do not succeed. In particular, I argue that Gustafsson faces a dilemma: either he abandons the principle of value–preference symmetry or he cannot make conceptual room for multiple permissible preferences.

Information

Type
Research Article
Information
Utilitas , Volume 28 , Issue 3 , September 2016 , pp. 239 - 253
Check for updates
Copyright
Copyright © Cambridge University Press 2015 

INTRODUCTION

We often compare different items in terms of value. Typically, we judge that one item is either better than, or equal to, or worse than another. Sometimes, however, we may judge that two items are value incomparable. Let us call these the standard value relations. According to some philosophers, these relations do not exhaust the conceptual space of value relations. Indeed, some think that there exists a fourth positive value relation, i.e. parity, which holds between two items when none of the standard value relations holds.Footnote 1 Let us call value relations such as parity non-standard value relations.

How can we analyse standard and non-standard value relations? One strategy consists in taking traditional analyses of non-comparative value and extending them to value relations. In recent years, several authors have been attracted by the Fitting Attitude-analysis (henceforth FA-analysis) of value.Footnote 2 In the non-comparative case, the FA-analysis holds that to judge that an item is good is to judge that that item is a fitting object of a favouring attitude. The concept of goodness is thus analysed in terms of a normative component, captured by the notion of ‘fittingness’, and a psychological component, captured by the notion of ‘favouring’. Assuming that this analysis provides a correct account of non-comparative value, how can we extend it in such a way as to provide an account of standard and non-standard value relations?

In a series of recent articles, Wlodek Rabinowicz has proposed an FA-analysis of comparative value that has three main features.Footnote 3 The first is what we may call the generic FA-thesis about comparative value. It says that comparative value judgements are equivalent to normative assessments of preferences. The second is a thesis about the normative component of the analysis. According to it, normativity has two different levels, i.e. the level of permissibility and the level of requiredness. The third is a thesis about the favouring component of the analysis. According to it, the only preference relations that exist are the standard preference relations, namely, strict preference, indifference, strict dis-preference, and – Rabinowicz adds – the preferential gap relation. Rabinowicz's idea is that we can provide an analysis of both standard and non-standard value relations in terms of the two levels of normativity and the standard preference relations just mentioned.

As will be illustrated below, Rabinowicz's framework has room for more value relations than preference relations. Yet, this feature of his account has not convinced everyone. In a recent article, Johan Gustafsson has argued that value and preference relations are symmetric.Footnote 4 According to Gustafsson, for every value relation, there is a corresponding preference relation, and vice versa. Gustafsson calls this the principle of value–preference symmetry. Rabinowicz's framework seems to violate this principle, since it does not appeal to non-standard preference relations in order to account for non-standard value relations. Gustafsson proposes to replace Rabinowicz's analysis with an alternative FA-analysis of comparative value, according to which all value relations (standard and non-standard) are analysed in terms of the corresponding (standard and non-standard) preference relations and one level of normativity only.

Gustafsson's idea is that, if the principle of value–preference symmetry holds, then, ceteris paribus, his analysis should be preferred to Rabinowicz's analysis. As we shall see in the next section, Rabinowicz's analysis associates each standard value relation to a standard preference relation. It is only when non-standard value relations are concerned that the symmetry between values and preferences breaks down. However, if the principle of value–preference symmetry holds, then, according to Gustafsson, considerations of simplicity and internal coherence suggest preserving the symmetry of the analysis also in the case of non-standard value relations. At the same time, these considerations decisively favour Gustafsson's analysis only if everything else remains equal, that is, only if Gustafsson's analysis is capable of accounting for all the phenomena accounted for by Rabinowicz's analysis. This explains the need to impose a ceteris paribus clause.

Gustafsson offers two arguments in support of his FA-analysis. The first is meant to show that the principle of value–preference symmetry is sound. The second is meant to show that the ceteris paribus clause is satisfied. More specifically, Gustafsson's first argument is that all the reasons for and against the existence of non-standard value relations are equally cogent reasons for and against the existence of non-standard preference relations. Hence, if there are non-standard value relations, there are also corresponding non-standard preference relations, as the principle of value–preference symmetry maintains. Gustafsson's second argument is that his account can make room for the idea that originally motivated Rabinowicz's account, namely, that, in some cases, different preference relations between a pair of items are equally permissible.

My goal in this article is to show that Gustafsson's arguments against Rabinowicz's FA-analysis do not succeed.Footnote 5 Before discussing these arguments, however, I shall present Rabinowicz's and Gustafsson's FA-analysis of comparative value in more detail.

RABINOWICZ’S VERSUS GUSTAFSSON’S FA-ANALYSIS OF VALUE RELATIONS

As stated in the introduction, Rabinowicz's FA-analysis of value relations is based on the distinction between two levels of ‘fittingness’, i.e. the level of ‘fittingness’ qua ‘permissibility’ and the level of ‘fittingness’ qua ‘requiredness’.

In fact, Rabinowicz's analysis of standard value relations involves only the level of requiredness. According to Rabinowicz, for any options x and y:

  1. (B) x is better than y if and only if one is required to prefer x to y.

  2. (W) x is worse than y if and only if one is required to prefer y to x.

  3. (E) x is equally good as y if and only if one is required to be indifferent between x and y.

  4. (I) x is value incomparable to y if and only if one is required to have a preferential gap between x and y.

The second level of normativity, i.e. the level of permissibility, comes into play only to characterize non-standard value relations. Consider parity. As we have seen, two items are on a par when neither is better than the other, nor equally good, and yet the two are value comparable. Within Rabinowicz's FA-analysis, parity is characterized as follows:

  1. (PR) x is on a par with y if and only if it is both permissible to prefer x to y and permissible to prefer y to x.

Rabinowicz provides a more formal interpretation of this FA-analysis in terms of the intersection of permissible preference rankings. On the basis of this framework, Rabinowicz is able to distinguish fifteen atomic value relations.Footnote 6 For the present purpose, however, it will suffice to consider only his informal analysis and to take parity as the paradigmatic example of non-standard value relation.

In contrast with Rabinowicz, Gustafsson thinks that it is possible to analyse both standard and non-standard value relations in terms of a single-level notion of ‘fittingness’, which, for purpose of consistency with Rabinowicz's framework, Gustafsson takes to be equivalent to the notion of ‘requiredness’.

Gustafsson's analysis of standard value relations is identical to Rabinowicz’s. The difference concerns his account of non-standard value relations. Consider parity. In order to analyse the concept of parity within a framework including only one level of normativity, Gustafsson suggests expanding the set of preference relations and making room for an additional, non-standard preference relation, i.e. the preferentially on a par relation. Gustafsson proposes the following analysis.

  1. (PG) x is on a par with y if and only if one is required to hold x and y preferentially on a par.

This strategy generalizes. According to Gustafsson, for any value relation (standard and non-standard), there exists a corresponding preference relation (standard and non-standard), in terms of which we can analyse the former. Gustafsson thinks that this account provides a more plausible FA-analysis of comparative value. In the next sections, I shall examine his arguments in support of this claim.

GUSTAFSSON’S FIRST ARGUMENT

The starting point of Gustafsson's first argument is the observation that, when we consider the standard relations, there is a correspondence between value relations and preference relations. In other words, when standard relations are concerned, values and preferences are symmetric. Gustafsson claims that, if this symmetry breaks down, it must be because there exists some non-standard value relation that lacks a corresponding preference relation. In order to show that this is the case, one needs a cogent argument for a non-standard value relation, which is not, mutatis mutandis, a cogent argument for a non-standard preference relation. According to Gustafsson, this is where the problem begins, since all the arguments for and against the existence of non-standard value relations seem to be equally cogent, mutatis mutandis, as arguments for and against the existence of non-standard preference relations. If this is true, then we should conclude that value and preference relations are always symmetric, both in the standard and in the non-standard case. Since Rabinowicz's FA-analysis is incompatible with this conclusion, we have reason to prefer Gustafsson's framework to Rabinowicz’s.

What are the typical arguments in favour of non-standard value relations? The two main arguments discussed in the literature are the small improvement argument and the chaining argument. On the one hand, the small improvement argument is meant to show that there exist cases where, for some x and y, x is not better than y, x is not worse than y, and x is not equally good as y. On the other hand, the chaining argument is meant to show that such cases are not cases of value incomparability, i.e. x and y are not value incomparable. If both arguments are successful, it follows that there exists an additional value relation that holds in some circumstances between the items, namely, axiological parity.

Gustafsson thinks that these arguments can be reformulated in terms of preferences. If so – he claims – one important question arises. If, in the axiological case, the small improvement argument and the chaining argument are sufficient for concluding that there exists a fourth positive value relation, i.e. axiological parity, which cannot be reduced to the standard value relations, why are they not also sufficient, in the preferential case, for concluding that there exists a fourth positive preference relation, i.e. preferential parity, which cannot be reduced to the standard preference relations? Indeed, if the small improvement argument and the chaining argument apply to the preference case, then we have reason to think that value and preference relations are symmetric.Footnote 7

In a similar vein, Gustafsson thinks that the objections against the small improvement argument and the chaining argument can be reformulated in terms of preferences. In the axiological case, these objections are meant to show that there exists no independent relation of axiological parity, since all the alleged cases of axiological parity can be accounted for, in one way or another, in terms of the standard value relations. The question is: if these arguments are cogent in the axiological case, why should we not take them to be cogent also in the preferential case? That is, why should we not take them to show that there exists no independent relation of preferential parity? Once again, if the objections against the small improvement argument and the chaining argument apply to the preference case, then it seems that we have no reason to think that value and preference relations are asymmetric.

It must be stressed that Gustafsson is not interested in taking sides in the substantive debate about parity. What matters to him is just that both the arguments in favour of the existence of parity and those against it apply to the preferential case as well as to the axiological case. In fact, Gustafsson's argument for the principle of value–preference symmetry is supposed to be independent of the outcome of the substantive debate about parity. Indeed, Gustafsson's point is simply that, if there exists a non-derivative relation of axiological parity, then there exists also a non-derivative relation of preferential parity (and vice versa); and if there exists no non-derivative relation of preferential parity, then there exists no non-derivative relation of axiological parity either (and vice versa).

DISCUSSION OF GUSTAFSSON’S FIRST ARGUMENT

At first sight, Gustafsson's argument seems convincing. In this section, however, I shall argue that his argument is more problematic than it initially appears. The first thing to observe is that it is not clear how exactly Gustafsson envisages the reformulation of the arguments in favour of and against the existence of parity in the preference case. There are two possibilities. On the one hand, these arguments can be reformulated in terms of actual preferences. On the other hand, they can be reformulated in terms of required preferences. Consider the first possibility. (For simplicity, I shall focus only on the arguments in favour of the existence of parity as a non-derivative relation.) When reformulated in terms of actual preferences, the small improvement argument is meant to show that there exist cases where, for some x and y, x is not strictly preferred to y, y is not strictly preferred to x, and x is not indifferent to y. In turn, the chaining argument is meant to show that such cases are not just cases where one has a preferential gap between x and y. Rather, they are cases where an additional preference relation holds between them, namely, preferential parity.

Gustafsson's central claim is that these arguments are as cogent as in the axiological case. However, there seems to be reason to doubt this claim. Let us take the chaining argument as an example. In the axiological case, the argument rests on the principle according to which, if two items x and y are comparable in terms of value and if a third item z differs from y only with respect to one dimension relevant for comparison, then x and z are also comparable in terms of value. Ruth Chang calls this the Small Unidimensional Difference Principle.Footnote 8 When reformulated in terms of actual preferences, the principle states that if no preferential gap holds between two items x and y and if a third item z differs from y only with respect to one dimension relevant for comparison, then no preferential gap holds between x and z. Contrary to what Gustafsson suggests, however, the two versions of the principle do not seem to be equally compelling. To see this, let us observe that, if the FA-analysis of standard value relations is true, then the axiological version of the Small Unidimensional Difference Principle can be equivalently reformulated in terms of required preferences. Indeed, the axiological version of the principle is formulated in terms of standard value relations and, according to the FA-analysis of standard value relations, the latter can be analysed in terms of required preference relations. If so, the axiological version of the Small Unidimensional Difference Principle is logically equivalent to the following version of the principle in terms of required preferences: if one is required to have no preferential gap between two items x and y and if a third item z differs from y only with respect to one dimension relevant for comparison, then one is required to have no preferential gap between x and z. Given this logical equivalence, in order for his argument to succeed Gustafsson must show that the version of the principle in terms of actual preferences is as cogent as the version in terms of required preferences. When we compare the two, however, we can easily see that they put forward claims of a very different sort, i.e. an empirical claim versus a normative claim. This should immediately cast some doubts on the claim that the two versions of the principle are equally compelling. Indeed, one has to back them with different kinds of considerations. Most importantly, there is no obvious reason to think that these considerations will give equally strong support to the two versions of the principle. In fact, it seems to me that the version of the Small Unidimensional Difference Principle in terms of required preferences is considerably more plausible than the one in terms of actual preferences.Footnote 9 While a conclusive assessment of the latter requires empirical studies concerning the behaviour of preferences in the face of small unidimensional changes, it is indeed not unreasonable to think that small unidimensional changes may often be sufficient to generate gaps in one's actual preferential attitudes, even when such gaps are (arguably) unjustified from a rational point of view. If this is true, the upshot is the following: since the chaining argument rests on the Small Unidimensional Difference Principle, it follows from the previous discussion that, when it is reformulated in terms of actual preferences, the chaining argument is not as cogent as in the axiological case.

Suppose, then, that the small improvement argument and the chaining argument are reformulated in terms of required preferences, rather than actual preferences. When modified accordingly, these arguments are meant to show that there exist some x and y such that one is not required to strictly prefer x to y, nor required to be indifferent between x and y, nor required to strictly prefer y to x and, yet, one is required not to have any preferential gap between them. Is Gustafsson's case in favour of the existence of preferential parity more convincing when the small improvement argument and the chaining argument are reformulated thus? Once again, there is reason to doubt that.

To begin with, it should be obvious by now that the possibility of such a reformulation is not at all surprising, since it is entailed by the FA-analysis of standard value relations, to which both Rabinowicz and Gustafsson subscribe. This remark serves to highlight one important point. The fact that the arguments in favour of axiological parity can be equivalently reformulated in terms of required preference relations does not force us to accept the conclusion that, if there exists a relation of axiological parity, there exists also a relation of preferential parity. In fact, it still leaves us with a choice between two alternative conclusions. On the one hand, we may conclude that cases of axiological parity are cases where a relation of preferential parity is required, as Gustafsson does. On the other hand, we may conclude that cases of axiological parity are cases where multiple standard preference relations are equally permissible, as Rabinowicz does.

There is indeed an important difference between the axiological and the preferential cases. In the axiological case, if we recognize that two items are not related in any of the standard ways, we have no choice but to conclude that these items are related in a non-standard way. Indeed, if neither the betterness, nor the equality, nor the worseness relations hold between two items, yet these items are still comparable in terms of value, it must be because there exists an additional, non-standard value relation that holds between them, i.e. the axiological parity relation. In the preferential case, by contrast, even if we admit that in some circumstances one is not required to have any unique standard preference relation, we are not obliged to conclude that one is thereby required to have a non-standard preference relation, i.e. the preferential parity relation. In fact, we may simply conclude that, in those circumstances, one is merely permitted to have different standard preference relations. If so, the question arises: is there any reason to favour the former conclusion, as Gustafsson recommends us to do?

Perhaps, one reason is the following. If we admit the existence of preferential parity, we obtain a more elegant FA-analysis of comparative value, according to which all value relations are analysed in terms of the corresponding preference relations. The problem is that elegance is but one consideration, which must be weighed against other theoretical considerations. One of these is parsimony. It seems clear, however, that Rabinowicz's FA-analysis of value relations is more parsimonious than Gustafsson’s, since it is capable of accounting for all value relations by appealing only to the standard preference relations. Moreover, as things stand, considerations of parsimony seem to be sufficient to trump considerations of elegance. After all, why should we postulate an unfamiliar preference relation, whose features remain utterly mysterious, simply because that makes our analysis of value relations more elegant? Occam's razor looms large.

In order to resist this conclusion, Gustafsson must say something more about preferential parity, so as to make this notion a little less obscure. It is worth observing that, in his article, Gustafsson tries to set this challenge aside by arguing that: ‘Explaining preference relations . . . falls outside the scope of a fitting-attitude analysis of value.’Footnote 10 Yet, this response is problematic. If we had an independent reason to believe the principle of value–preference symmetry, then we could accept an FA-analysis of axiological parity based on an undefined notion of preferential parity. However, whether or not to believe the principle of value–preference symmetry is precisely what is at stake. If so, leaving the notion of preferential parity unexplained is just a way of begging the question against Rabinowicz.Footnote 11

Gustafsson must thus provide a positive account of preferential parity. In particular, his account must satisfy at least three conditions. First, it must offer a characterization of preferential parity as an independent preference relation, i.e. a relation irreducible to the standard preference relations.Footnote 12 Second, it must not entail that, if one is required to hold two items preferentially on a par, then it is permissible for one to have opposite preferences between these items, and vice versa. If this were the case, then Gustafsson's FA-analysis of axiological parity would simply collapse into Rabinowicz's FA-analysis. Last, but not least, Gustafsson's account of preferential parity must not render his FA-analysis of value relations significantly less parsimonious than Rabinowicz’s.

These conditions set some significant constraints on the kind of account that Gustafsson can legitimately offer. To give just one example, the first condition prevents us from defining preferential parity as the disjunction of two (or more) standard preference relationsFootnote 13 or as a state of wavering between two (or more) standard preference relations.Footnote 14 The reason is that such definitions do not construe preferential parity as an independent preference relation. Clearly, nothing said so far shows that there exists no account of preferential parity that can satisfy all these conditions. However, the burden of the proof lies entirely on Gustafsson. Indeed, until we can make sense of the notion of preferential parity in a satisfactory way, we are justified in holding on to an analysis of non-standard value relations couched in terms of categories with which we are already familiar, e.g. the categories of standard preference relations and the categories of requiredness and permissibility. I conclude that Gustafsson's first argument is not persuasive.Footnote 15

GUSTAFSSON’S SECOND ARGUMENT

As we have seen, according to Rabinowicz, there exist some circumstances in which it is permissible for one individual to have different preference relations between a pair of items (although not simultaneously). Gustafsson recognizes this possibility. The problem is that his framework does not seem to allow for cases of multiple permitted preferences, since, for each value relation, there is a unique preference relation that one is required to have. If this is true, then Gustafsson's framework does not capture one important feature of the normativity of preferences, which Rabinowicz's framework does capture.

Gustafsson disagrees. He thinks that the conclusion that his framework cannot make room for the idea of multiple permitted preference relations stems from an illegitimate assumption, namely, that, for any two distinct preference relations a and b and any two items x and y, if one is required to have a preference relation a between x and y, then one is not permitted to have a preference relation b between x and y. Gustafsson thinks that this claim is mistaken and should be rejected.

Consider a situation where two items x and y are axiologically on a par. As we have seen, Gustafsson thinks that, in this case, one is required to hold x and y preferentially on a par. However, according to Gustafsson, it does not follow that it is not permissible for one to have also a standard preference relation between x and y.

Since neither ‘better’ nor ‘worse’ holds, one is not rationally required to prefer x to y and one is not rationally required to prefer y to x. Yet one might be both rationally permitted to prefer x to y and rationally permitted to prefer y to x since this is compatible with having the rationally required preference relation, that is, to hold x and y preferentially weakly on a par.Footnote 16

If this reasoning is correct, then Gustafsson's framework can make room for multiple permitted preference relations, like Rabinowicz's account.

DISCUSSION OF GUSTAFSSON’S SECOND ARGUMENT

In what follows, I shall argue that Gustafsson's second argument fails. To begin with, let us agree that, if two attitudes are mutually exclusive and if one is required to have one of these attitudes, then one is required not to have the other attitude. This can be expressed, more formally, by the following principle:

  1. (K) For any two preference relations a and b and any two items x and y, if one is required to have a preference relation a between x and y and if it is not possible for a and b to hold together, then one is required not to have a preference relation b between x and y.

Clearly, (K) entails that, if one is required to have a preference relation a between x and y and if it is not possible for a and b to hold together, then one is not permitted to have a preference relation b between x and y. In order to allow for cases of multiple permitted preferences, Gustafsson must thus admit the possibility that two preferential attitudes may not be mutually exclusive. In such cases, the following principle holds:

  1. (L) For any two preference relations a and b and any two items x and y, if one is required to have a preference relation a between x and y and if it is possible for a and b to hold together, then there may exist some cases where one is permitted to have a preference relation b between x and y.

The crucial question is whether there exist pairs of preference relations that are not mutually exclusive. Gustafsson claims, minimally, that the preferential parity relation is consistent with any of the standard preference relations. If he is right, then there may exist some cases in which it is possible to have both the required non-standard preference relation and another permissible standard preference relation.

Is Gustafsson right? The first thing to notice is that his solution seems ad hoc. Standard preference relations are typically considered to be mutually exclusive. If so, why should we think that the situation is any different when we consider non-standard preference relations? There seems to be no independent reason to think that standard and non-standard preferential attitudes are not mutually exclusive. Gustafsson would perhaps claim that standard and non-standard preferential attitudes are compatible when the latter are derived from the former, e.g. if preferential parity is simply an indeterminate standard preference relation.Footnote 17 Yet, this reply is problematic. As we have seen, if preferential parity exists as a genuine preference relation, then it exists as an irreducible preference relation. However, the definition of preferential parity under consideration presupposes the opposite, namely, that preferential parity is somehow reducible to the standard preference relations. Hence, this definition effectively rules out the existence of preferential parity as a genuine preference relation. The upshot is the following: either Gustafsson is forced to deny the existence of preferential parity from the start or his FA-analysis is incapable of accounting for multiple permitted preferences. Neither option is, of course, especially appealing.

Perhaps, this criticism is unfair. After all, Gustafsson claims to be concerned not with strict parity, but with weak parity. In other words, Gustafsson only claims that it is possible for an individual to hold two items preferentially weakly on a par and to have either a strict preference between them or an attitude of indifference. To assess this claim, we need to understand what the weak parity relation is. Unfortunately, Gustafsson is not particularly clear on this subject. In his text, he simply claims that weak parity should be understood as a pre-theoretical notion. After that, he offers some examples of axiological weak parity. Gustafsson thinks, for instance, that two items x and y may be axiologically weakly on a par when x is slightly better than y.Footnote 18 Thus, in contrast with the theoretical notion of axiological strict parity, which holds when none of the standard value relations holds, the pre-theoretical notion of axiological weak parity may hold together with a standard value relation such as betterness – at least if we assume that slight betterness entails betterness.Footnote 19 If this is true, then, in a similar way, the preferential weak parity relation may hold together with the strict preference relation.

However, this line of thought presents at least two problems. The first is that the compatibility between the preferential weak parity relation and the standard preference relations is simply assumed, rather than demonstrated. The second is that, to the extent that we take the claim that weak parity is a pre-theoretical notion at face value, then Gustafsson's argument can immediately be discarded. The reason is that the FA-analysis is concerned with theoretical, not pre-theoretical, value relations. So, this line of thought is precluded from the start.Footnote 20

Suppose, then, that we go back to the theoretical understanding of axiological (strict) parity as a relation that holds when none of the standard value relations holds. In this case, there exists another and more serious reason to think that Gustafsson's argument is ultimately unsuccessful. As we have seen, Gustafsson's argument depends on the claim that it is possible for an agent to hold a standard and a non-standard preferential attitude together. Suppose, for the sake of the argument, that this is actually the case. One important result follows: the symmetry between preference relations and value relations breaks down. Consider the case of parity. As far as preferences are concerned, Gustafsson maintains that an individual may hold two options x and y preferentially on a par and, at the same time, (e.g.) strictly prefer x to y. From an axiological point of view, however, cases of parity are cases where none of the standard value relations holds. This means that the axiological parity relation cannot hold together with the betterness relation. If this is true, then value and preference relations behave rather differently. While the individual can have two distinct preference relations between a pair of items, the corresponding value relations between these items cannot hold together. This constitutes a violation of the principle of value–preference symmetry. Indeed, suppose we define a new preference relation c as the conjunction of the preferential parity relation and the strict preference relation.Footnote 21 We can immediately see that the principle of value–preference symmetry no longer holds, since there exists no value relation corresponding to c. Hence, Gustafsson faces a dilemma: either he abandons the principle of value–preference symmetry or he cannot make conceptual room for multiple permissible preferences. Either way, Gustafsson is in trouble.Footnote 22

References

1 See Chang, R., ‘Introduction’, Incommensurability, Incomparability, and Practical Reason, ed. Chang, R. (Cambridge, MA, 1997), pp. 1–34 Google Scholar; and Chang, R., ‘The Possibility of Parity’, Ethics 112 (2002), pp. 659–88CrossRefGoogle Scholar.

2 See Scanlon, T., What We Owe To Each Other (Cambridge, MA, 1998)Google Scholar; Rabinowicz, W. and Rønnow-Rasmussen, T., ‘The Strike of the Demon: On Fitting Pro-Attitudes and Value’, Ethics 114 (2004), pp. 391–423 CrossRefGoogle Scholar; Danielsson, S. and Olson, J., ‘Brentano and the Buck-Passers’, Mind 116 (2007), pp. 511–22CrossRefGoogle Scholar.

3 See Rabinowicz, W., ‘Value Relations’, Theoria 74 (2008), pp. 18–49 CrossRefGoogle Scholar; Rabinowicz, W., ‘Value Compared’, Polish Journal of Philosophy 3 (2009), pp. 73–96 CrossRefGoogle Scholar; Rabinowicz, W., ‘Value Relations Revisited’, Economics and Philosophy 28 (2012), pp. 133–64CrossRefGoogle Scholar. Rabinowicz develops, and substantially modifies, a line of thought originally proposed by Joshua Gert (see Gert, J., ‘Value and Parity’, Ethics, 114 (2004), pp. 492–510 CrossRefGoogle Scholar). Since Gert's proposal is subject to some important objections (see Chang, R., ‘Parity, Interval Value, and Choice’, Ethics 115 (2005), pp. 331–50CrossRefGoogle Scholar, at 337–44; Rabinowicz, ‘Value Relations’, pp. 30–6), in what follows I shall focus only on Rabinowicz's FA-analysis.

4 Gustafsson, J., ‘Value–Preference Symmetry and Fitting-Attitude Accounts of Value Relations’, The Philosophical Quarterly 63 (2013), pp. 476–91CrossRefGoogle Scholar.

5 In this article, I revise and extend my discussion of Gustafsson's arguments in Rossi, M., ‘Sur la symétrie présumée entre valeurs et préférences’, Les ateliers de l’éthique/The Ethics Forum 9 (2014), pp. 82–98 Google Scholar.

6 It is worth noticing that, on Rabinowicz's approach, parity turns out to be a non-atomic type of value relation. In fact, within Rabinowicz's framework, parity is the collection of four atomic types.

7 In further support of this line of thought, Gustafsson adds that the authors who first proposed the small improvement argument and the chaining argument (i.e. respectively, de Sousa, R., ‘The Good and the True’, Mind 83 (1974), pp. 534–51CrossRefGoogle Scholar, at 544–5; and Chang, ‘The Possibility of Parity’, p. 666) have either formulated these arguments in terms of preferences, rather than value, or explicitly claimed that their arguments apply to preference relations as much as to value relations.

8 See Chang, ‘The Possibility of Parity’, pp. 674–5; see also Gustafsson, ‘Value–Preference Symmetry’, p. 484.

9 This is not to say that the version of the Small Unidimensional Difference Principle in terms of required preferences is fully convincing. For instance, one may argue against it by claiming that, in cases of small unidimensional changes, one may sometimes be permitted (although not required) to have a preferential gap between the items. Notice that, if one endorses this line of thought, one will also be logically compelled to endorse a weaker axiological version of the principle, stating that, if two items x and y are fully comparable in terms of value and if a third item z differs from y only with respect to one dimension relevant for comparison, then x and z are weakly comparable in terms of value; where two items x and y are fully comparable if and only if one is required to prefer one of these items to the other or to be indifferent between them; and where two items x and y are weakly comparable if and only if one is not required to have a preferential gap between them. For the notions of full and weak comparability, see Rabinowicz, ‘Value Relations’, pp. 27–8.

10 Gustafsson, ‘Value–Preference Symmetry’, p. 488.

11 One may be tempted to argue that, if we cannot make sense of preferential parity, then we cannot make sense of axiological parity either. Once again, however, this would be true only if we had independent reason to believe the principle of value–preference symmetry. Since we do not, and since there exists an alternative way of making sense of the notion of axiological parity, i.e. Rabinowicz's, then we should simply conclude that, in some cases, there exist value relations that lack corresponding preference relations. In other words, unless Gustafsson can provide a suitable account of preferential parity, we have no reason to think that value and preference relations are symmetric.

12 The reason for this condition is the following. As we have seen above, if axiological parity exists as a genuine positive value relation, then it exists as a relation that cannot be reduced to the standard value relations. By the principle of value–preference symmetry, it follows that, if preferential parity exists as a genuine positive preference relation, then it exists as a relation that cannot be reduced to the standard preference relations.

13 Incidentally, this proposal would also be incompatible with another principle to which Gustafsson subscribes in his article, namely, the principle of Non-irrationality of axiological constrainedness. According to it, if none of a set of value relations holds between x and y, then one is not rationally required to have one of the corresponding preference relations. Since axiological parity holds when none of the standard value relations holds, the principle of Non-irrationality of axiological constrainedness entails that, in such cases, one is not required to have any of the standard preference relations. However, if we define preferential parity as the disjunction of two (or more) standard preference relations, it precisely follows that, when axiological parity holds, one is required to have either one of the corresponding standard preference relations – a clear violation of the principle of Non-irrationality of axiological constrainedness.

14 See W. Rabinowicz, ‘From Values to Probabilities’, forthcoming in Synthese, fn. 11.

15 Wlodek Rabinowicz has suggested to me that there exists a third way of reconstructing Gustafsson's argument, i.e. in terms of permissible, rather than actual or required, preferences. When reformulated accordingly, the small improvement argument states that there exist some x and y such that one is not permitted to strictly prefer x to y, nor permitted to be indifferent between x and y, nor permitted to strictly prefer y to x. In turn, the chaining argument states that, in such cases, one is permitted not to have any preferential gap between x and y. The combination of these arguments is incompatible with Rabinowicz's analysis, but compatible with Gustafsson's. The question, then, is whether the small improvement argument and the chaining argument, thus formulated, are as cogent as in the axiological case. In fact, there is reason to doubt that. Let us focus on the small improvement argument. Recall that the axiological version of this argument can be equivalently reformulated in terms of required preferences. When reformulated accordingly, the small improvement argument states that there exist some x and y such that one is not required to strictly prefer x to y, nor required to be indifferent between x and y, nor required to strictly prefer y to x. The important thing to notice is that, in this case, it may still be permissible for one to have any of these preference relations between x and y. By contrast, this is excluded by the version of the small improvement argument in terms of permissible preferences. For this reason, the latter appears to be a considerably more demanding, and less plausible, argument.

16 Gustafsson, ‘Value–Preference Symmetry’, p. 490.

17 As Gustafsson suggests elsewhere. See Gustafsson, J., ‘Indeterminacy and the Small-Improvement Argument’, Utilitas 25 (2013), pp. 433–45CrossRefGoogle Scholar, at 445.

18 Another possibility is that two items x and y may be axiologically weakly on a par when x is equally good as y.

19 Notice, however, that Gustafsson allows also for the possibility that axiological weak parity may hold when none of the standard value relations holds. See Gustafsson, ‘Value–Preference Symmetry’, p. 490.

20 Arguably, the FA-analysis is a theoretical attempt to clarify, refine and systematize our pre-theoretical understanding of value relations. In fact, the latter seems to be too ambiguous and confused for it to serve as the main evidence in favour of a theoretical principle such as (L).

21 Accordingly, we can say that an individual has a preferential attitude c between two items x and y if and only if she both holds x and y preferentially on a par and strictly prefers x to y.

22 I am especially grateful to Wlodek Rabinowicz for his very useful comments as a no-longer-anonymous referee for this article. I would also like to thank Krister Bykvist, Erik Carlson, Denis Courville, Jens Johansson, Christopher Kelly, Sebastiano Moruzzi, Graham Oddie, Niklas Olsson-Yaouzis, Michele Palmira, Jean-Charles Pelland, Daniel Ramöller, Frans Svensson, Jennifer Szende, Christine Tappolet, Giorgio Volpe and one anonymous referee for their helpful comments on previous versions or presentations of the article.