Moral subjectivism and the semantics of disagreements 1 Subjetivismo moral e a semântica dos desacordos

In this paper, I discuss which semantic theory moral subjectivists should adopt. Moral subjectivism is understood broadly to include all theories according to which moral sentences are truth-apt, at least sometimes true, and made true by the mental attitudes of certain relevant agent or set of agents. Due to the breadth of this definition, an initial concern is whether a unified semantic approach is able to accommodate all varieties of subjectivism. I argue that it is. I then proceed to analyse the main semantic theories for moral sentences as they apply to the standard issue of moral disagreements. I conclude in favour of so-called Non-Indexical Contextualism


3/11
theories, the relevant agent is not the actual agent who is entertaining a certain moral discourse or practice.It might be a divine being or an idealised counterpart of the actual agent.But the point is that, for any given moral judgment or behaviour, all subjectivist theories contend that there is a fixed relevant agent or set of agents, whose (real or virtual) verdict establishes what ought to be done.For ease of presentation, from now on I will proceed as if the definition of what ought to be done by any particular agent is set by the moral standards of that very agent.Nonetheless, with appropriate modifications, the same points apply to other varieties of moral subjectivism.
The question I would like to investigate next is: What is the most appropriate semantic theory for moral subjectivism?By semantic theory I mean "a theory which assigns semantic contents to expressions of a language" (Speaks, 2021).It is standard to assume that human language and communication manifest both semantic and pragmatic features.Although the semantics/pragmatics cleavage is by no means uncontroversial, I will leave it at an intuitive level and focus my concerns here on the semantics of moral thought and talk. 4Is there a single and convincing way of attributing content and truth-value to moral expressions that can accommodate the variety of subjectivist positions alluded to above?Or is each kind of moral subjectivism better served by a specific kind of semantic theory?
Call invariantism the view that the content of moral judgments does not vary according to context.Whenever a moral judgment is issued, what is said according to this view is fixed before any features of the context of utterance or the context of assessment are taken into consideration.Some subjectivist views, such as Korsgaard's, are compatible with an invariantist moral semantics.Recall that, for her, what an agent has moral reason to do formally follows from her practical standpoint and this standpoint is not characterized by any contingent values this agent might endorse, but rather stems for the appropriate application of her practical reason to the relevant elements that make up the circumstances of her deliberation.The agent's moral duties will be acknowledged by anyone -including the agent herself -who is able to follow the rational path from her circumstances to her practical decision via the application of the universally valid principles of practical reason, foremost among them the Categorical Imperative.If that is the case, then, the meaning and truth of moral judgments as uttered by any rational agent will be always fixed, so long as the conditions are themselves fixed.
It seems that other, non-universalistic versions of subjectivism, such as Street's or Sobel's, cannot help themselves to this invariantist conception of the semantics of moral judgments.According to them, what an agent has moral reason to do will at least partially reflect contingent facts about that agent, so that different, but equally situated agents may have different moral reasons.If that is the case, then an invariantist semantics will not necessarily suffice to interpret the moral judgments of different agents.It may be that the contents of a given agent's moral judgments are partly constituted by that very agent's moral standards.
On a first pass, then, this suggests that part of the dispute within subjectivism between those favouring a universalistic approach and those opposing it is fought in semantic terms.Which of these approaches is more plausible would (partially) boil down to which semantic theory is more independently robust: invariantism or indexical contextualism (as the opponent in this case is often called -more on this below).
That is possible, but I would like to propose that it is not the most fruitful framing of the dispute.To see this, notice that there are certain versions of invariantism and contextualism that are capable of accommodating the different subjectivist views we have been considering.We will go through these versions in more detail in the next section.For now, take first the semantic theory according to which the content of an agent's moral judgments is (partially) determined by that agent's moral standards.This is the view we called indexical contextualism above.This view seems to be at odds with the universalistic subjectivism of, say, Korsgaard.On her view, what moral agents can correctly state does not vary according to the potentially contingent moral standards of the agents.On the contrary, all rational agents are under the same moral obligations.But the conflict is merely apparent.It is perfectly possible both to hold that the content of moral judgments is (partially) determined by the agent's moral standards and to claim that there is only one, universally correct moral standard.All this implies is that, under equal circumstances, for any two rational agents, the contents of their moral judgments will necessarily be the same, because their moral standards, which (partially) fix the content of the judgments, will also be the same.In other words, the contextual features that contribute to the fixation of the content of the judgment (i.e., the agents' moral standards) will coincide and, as a consequence, so will the contents of their (correct) judgments.So it is perfectly possible to support universalistic moral subjectivism and at the same time defend an indexical contextualist view about the semantics of moral judgments.
Nor are all invariantist theories of moral semantics unavailable to moral subjectivists who reject the universalistic ambition described above.For it is possible to hold that, although the content of different moral agents' judgments are the same under equal conditions (as invariantism would have it), the truth value of those same judgments may vary.So the more relativistically inclined moral subjectivist may wish to preserve an invariantist semantics, according to which different agents with different moral standards do not utter judgments with different contents under equal conditions, but at the same time claim that the truth of the judgments ought to be assessed according to the (different) moral standards of the concerned agents. 5If that is so, an invariantist semantics will be available for subjectivists of both a universalistic and a non-universalistic variety.
Having said that, what I have set out to do in the remainder of this text is to go through the main contenders for best theories of moral semantics, under the presupposition of moral subjectivism.As we have seen, mere compatibility with the different sorts of moral subjectivism will not be enough to favour one theory over the others (though it will play a minor part towards the end of the article).So it will be necessary to engage with the semantic dispute on its own terms, so to speak, and search for the relative merits of the contending approaches.As per usual, I focus on the interpretation of moral disagreements, for they allow us to best tease out the relevant features of the different theories.I will be analysing the prospects for indexical contextualism, non-indexical contextualism, and assessor relativism.The latter two views are invariantist in the sense discussed above, while the former is not.My conclusion will be that non-indexical contextualism has the edge over the other positions, but that requires some argumentation.
Before that, one final caveat.As we saw, universalistic subjectivism is compatible with both invariantist and indexical contextualism semantics.However, because of its universalistic ambitions, it predicts that each and every instance of moral disagreement among rational agents will be fraught with mistakes.That is because, had the agents reasoned properly and holding constant the access to pertinent evidence, they would have reached the same verdict.For that reason, in what follows I will concentrate the discussion on relativistic subjectivist theories, since that is where the most interesting and theoretically divisive cases of moral disagreement occur.

The Semantics of Moral Disagreements
Some philosophers believe that there is no such thing as (intercultural) moral disagreement.6However, although highly stylized, there seems to be nothing impossible about a disagreement, say,

5/11
between an average endorser of Enlightenment Morality and Gibbard's Ideally Coherent Caligula 7 .If that is the case, accounting for such disagreement is one of the things we are entitled to expect from a moral theory. 8o how do we explain moral disagreements with a little more detail?Most authors traditionally called relativists 9 in metaethics can be classified as what is now called, under the influence of the philosophy of language, indexical contextualists (IC). 10In its simplest form, moral indexical contextualism is just like contextualism for indexical terms like "today".So, if I say, on 5 th October 2015: 3) Today is sunny.
And you say, on 5 th October 2016: 4) Today is not sunny.
We are not disagreeing so much as we are talking about different things.If we substitute "today" in 3 for "5 th October 2015" and "today" in 4 for "5 th October 2016" it becomes immediately clear that 3 and 4 are not incompatible assertions.According to IC about morality, the same happens with moral terms.So when Immanuel says: 5) Lying is always wrong.And Jeremy says: 6) Lying is not always wrong The terms "wrong" in 5 and "wrong" in 6 mean different things.On one interpretation, they are abbreviations for "wrong according to Immanuel's standard" in 5 and "wrong according to Jeremy's standard" in 6.It becomes clear, then, that Immanuel and Jeremy are not engaged in asserting two mutually excluding sentences.Rather, their assertions are perfectly compatible with one another and both might very well be correct at the same time.
This theory is perfectly suitable for so-called relativists who claim that moral disagreement is impossible, as mentioned before.Other defenders of IC, who wish to preserve the possibility of cross-cultural disagreements, will seek to account for them in pragmatic terms, appealing, e.g., to the notion of metalinguistic negotiations or to other kinds of non-doxastic disagreements. 11Although there are multiple strategies available for the IC, it seems reasonable to suppose that, at least sometimes, disagreements are doxastic, i.e., about conflicting contents.Therefore, I focus on this paper on the semantic aspects of disagreements, the reality of pragmatic features notwithstanding.Referring specifically to the contents asserted by the disagreeing parties, then, let me present some problems for indexical contextualism with a little bit more detail.
First, there is a problem we might call the problem of the loss of normativity.When X says "Child abuse is wrong" and Y says "Child abuse is not wrong" they are both uttering sentences with normative content.When the indexical contextualist translates X's sentence to "Child abuse is wrong according to X's standards" the translated sentence is not a normative sentence anymore, but rather merely a descriptive sentence about the components of X's standard or a merely logical claim about what follows from X's moral parameters.As Boghossian clearly expressed, on this view we seem unable to explain how there could be such a thing as a disagreement about normative matters (Boghossian, 2011, p. 58).
Another problem for the IC, also diagnosed by Boghossian, is that, on this picture, 7 The Ideally Coherent Caligula (ICC) is a fictional character created by Gibbard to mobilize intuitions against constructivist views such as Korsgaard's.The ICC's only policy in life is to maximize the suffering of others, which he pursues with the utmost coherence Gibbard (1999, p. 145). 8(Street, 2009) also acknowledges the importance of accounting for this kind of case. 9In fact, not only relativists.As we saw in the previous section, subjectivists, including those with universalistic ambitions, can be accommodated within an indexical contextualist approach to semantics. 10For an introduction to the discussion that follows, see Stojanovic (2017).
11 See (Zeman, 2017) for an overview of contextualist strategies.See (Plunkett & Sundell, 2013) for the notion of metalinguistic negotiations.A predecessor of this notion is to be found in (Chalmers, 2011).

6/11
there is nothing very exciting going on as far as truth is concerned.Contents have absolute truth-values.The only sense in which anything is relative to anything else is that the thinkable content expressed by a token of a sentence type is relative to that token's context of utterance (Boghossian,  2011, p. 59).
In other words, on this view, the content of all moral claims is relative, but their truth-values are always absolute.Assessors should refer to the context of utterance of a given token of a sentence to assess its truth-value and all assessors should agree on the verdict.This is another way of saying that disagreement about content is impossible on indexical contextualism.Being so, this justifies looking for alternative views.
A closely related view allows for doxastic disagreement.According to Non-Indexical Contextualism (NIC) 12 , when Immanuel asserts 5 and Jeremy assert 6 they are genuinely disagreeing.On this view, Immanuel and Jeremy hold opposing views about the same proposition.So, while in IC, the content of the proposition is contextually defined in reference to the speaker's standard, in NIC the content of the proposition is fixed.So if Immanuel asserts 5 and Jeremy asserts the opposite of 5 (=6), they are talking about the same content.In the words of Berit Brogaard, "as the contents of 'right' and 'wrong' are context invariant, there is something for the disputants to disagree about, viz the invariable content the truth-value of which depends on the standards of the judge" (2008, p. 393). 13IC allows for doxastic disagreement and, at the same time, claims that both parties in the disagreement might be correct.When this happens, they are engaged in what might be called a faultless disagreement. 14The disagreement can be faultless because the standard of evaluation for each party's assertion is different.So it might very well be the case that, according to Immanuel's standard, 5 is justified, while according to Jeremy's standard 5 is not justified.When that is the case, an assertion of 5 by Immanuel and a denial of 5 by Jeremy constitute a faultless disagreement.According to this view, each agent's perspective does not step in at the moment of determining the content of what is asserted, as in IC.Rather, the agent's perspective is the function assigning truth-values to propositions (Kölbel, 2004,  p. 70).Assertions of the same proposition by different agents may have different truth-values.
Several authors have resisted the idea that disagreements such as the one just described could be coherently called faultless (see Boghossian, 2011; Richard, 2008; Wright, 2008).They all more or less point to a disquotational schema.The argument goes as this: assume that non-indexical contextualism is right and that the truth of normative assertions is only relative truth (truth relative to the perspective of the utterer).Even then, these critics claim, it must be valid of someone that judges p that she can say: "It is true that p".But if she can say "It is true that p", she can also say "It is false that not-p".But if she can say "It is false that not-p" she cannot coherently claim that anyone judging that not-p is not making a mistake.Hence the impossibility of faultless disagreement.
Commenting on an alleged faultless disagreement between D and N and referring to a group of views (Alethic Relativism) of which NIC is one specimen, once more Boghossian lays out the argument with precision:

7/11 (13) The content (p) is at best relatively true. (Alethic Relativism) (14) If D judges validly that p, it will also be valid for D to judge that It's true that p. (Truth is Disquotational within a perspective) (15) If D judges that It's true that p then D must, on pain of incoherence, judge that It's false that not-p. (16) If D judges that It's false that not-p, then D must, on pain of incoherence, judge that anyone who
judges not-p (e.g., N) is making a mistake.Therefore, (17) D must judge that N is making a mistake and so cannot regard the disagreement with N as faultless.Therefore,  (18) The disagreement between D and N is not faultless.(Boghossian, 2011, p. 62)   The argument is powerful but I believe it can be adequately countered.I will point at two alternative responses, without the ambition of having exhausted the matter.The first alternative, due to Berit Brogaard ( 2008), contradicts directly the argument.The point is to recall that this argument is presupposing that the default judge of a given truth-predicate is always the evaluator and not the speaker, as non-indexical contextualism would suggest.For, remember, what the example above shows is that D cannot simultaneously judge that not-p is false and judge that N is not making a mistake in asserting not-p.Brogaard's argument is to recall that, according to perspectivalism (her version of non-indexical contextualism), the truth-predicate means something different when it is restricted to the meta-linguistic level from what it means when it is not so restricted.When it is not so restricted, we can infer 'John is a firefighter' from '"John is a firefighter" is true'.But we cannot do this when it is restricted.When it is restricted, the utterance 'John is a firefighter' may be true if John is a firefighter from 1990 to 2000.So 'The utterance "John is a firefighter" is true' does not entail that John is a firefighter.In other words, some difference must exist between meta-linguistic uses of sentences where the truth-predicate is restricted to the meta-linguistic level and uses where it is not so restricted.(Brogaard, 2008, p. 407)   And she next claims, that the truth-predicate that is restricted to the meta-linguistic level is the usual one (given Kaplan-style semantics); s as uttered in c is true simpliciter iff the proposition expressed by s in c is true at the circumstance of evaluation determined by the context of the speaker who uttered s (and not the context of the semanticist).In other words, the main difference between [assessor] relativism and perspectivalism concerns meta-linguistic uses of sentences where the truth-predicate is restricted to the meta-linguistic level.(2008, p. 407-8)   The details of Brogaard's view are complex and need not detain us here.Her general point is that the validity of the disquotational schema that translates from a judgment that p to a judgment that p is true and, thus, grounds the whole argument of the critics of non-indexical contextualism is based on a presupposition that the truth-predicate in the analysed cases is not restricted to the meta-linguistic level.And, contrary to that, she claims that restricting the truth-predicate to the meta-linguistic level is the usual attitude within the framework of standard Kaplanian semantics.
Another, alternative approach to the criticism above is to accept it, but reduce significantly its import.The idea is to acknowledge that, for disquotational reasons, it is inappropriate, in the example above, for D to say that N's assertion of not-p is true.However, the suggestion now is to say that refraining from granting your interlocutor's assertion the status of "true" (or, mutatis mutandis, "false") doesn't amount to claiming that she is making a mistake.So, perhaps the notions of truth and falsehood cannot be used the way non-indexical contextualists originally thought they could, but surely some other no-tion, like accuracy, could do the work.In this proposal, D is unable to say that N's assertion that not-p is true, because D's use of the word "true" picks out necessarily D's standard, but D is able to say that N's assertion that not-p is accurate, if we postulate that accuracy is meant to pick out what is true according to the standard of the speaker, not the evaluator.Thus, if both D and N are able to see each other as accurate when asserting respectively p and not-p, then it seems like it is after all appropriate to speak of faultless disagreement, just like non-indexical contextualists recommend. 15ow, non-indexical contextualism is the view I endorse.However, it is worth briefly mentioning another alternative theory and sketching an argument for not embarking on it.In a series of papers (MacFarlane, 2005, 2007, 2009), culminating in his 2014 book, John MacFarlane has advanced the view he called Assessor Relativism (AR).Remember that IC arguably has problems with accounting for disagreements.For MacFarlane, NIC does not fare much better in this respect and this is so for reasons different from the ones we discussed so far.
Unlike IC, NIC accounts for the kind of disagreement that exemplifies what MacFarlane calls the condition of Noncotenability, i.e., the idea that I disagree with someone's doxastic attitude if, in order to incorporate her attitude, I would have to change my mind (2014, p. 121).However, MacFarlane claims there is a further sense of disagreement that NIC and IC are both not capable of accommodating.That is a disagreement fulfilling the condition of Preclusion of Joint Accuracy: "The accuracy of my attitudes (as assessed from any context) precludes the accuracy of your attitude or speech act (as assessed from that same context)" (2014, p. 129).In NIC, joint accuracy is not precluded.It is possible for A to be accurate in claiming that p, while B is accurate in claiming that not-p.Regardless of whether A, B or a third party assess A's and B's claims, both claims may be deemed accurate from the point of view of the same assessor.
Contrary to this, AR endorses the Preclusion of Joint Accuracy.For AR, the accuracy of a claim is always evaluated from the point of view of the assessor.So, to repeat the example above, if A claims that p and B claims that not-p, it is impossible for both to be accurate.For, if A is the one assessing the accuracy of both claims, A will do so with reference to her perspective.If, on the other hand, it is B who is assessing both claims, then B will do so with reference to his perspective.The perspective of reference is always the perspective of the assessor and, therefore, it is impossible for joint accuracy of incompatible claims to obtain.
Fully responding to this challenge would require detailed argumentation.I only sketch three arguments for sticking to NIC in the face of MacFarlane's critique.First, I am not fully convinced of the relevance of Preclusion of Joint Accuracy.It seems to me that Noncotenability captures the essence of what is going on in moral disagreements, from the semantic point of view.That is, Noncotenability captures the idea that parties in a disagreement express or hold opposing attitudes towards the same content and it is not obvious why we want anything more than that. 16econdly, assessor-relativism adds an extra layer of semantic complexity.As Brogaard notes, assessor relativism "relativizes sentence truth to a context of use and a context of assessment.As a result, what is said on a particular occasion does not have an absolute truth-value at the context of use."In 15 This final proposal might help the non-indexical contextualist in yet another way.Allegedly a problem for non-indexical contextualism comes from data from experimental philosophy roughly showing that people use words like "false" and "no" when opposing statements made by interlocutors whose standards of evaluation predict the opposite of what their standards (of the ones who react saying "false" and "no") would predict.By allowing faultless disagreement to coexist with coherent and justified attributions of falsehood to the assertions of the disagreeing party, non-indexical contextualism comes up as not obviously contrary to the available empirical evidence.See (Khoo & Knobe, 2018) for the empirical study and a defense of indexical contextualism based on data such as the above. 16I conjecture that wanting more of disagreements, that is, demanding that the opposing party cannot also be accurate moves beyond simply being in a state of disagreement and goes into the territory of what parties might be willing to do when engaged in a disagreement.This is the topic for a paper on the pragmatics of moral disagreements, not the semantics.See (Cappelen &  Hawthorne, 2009) for the notion of disagreements in state, as opposed to disagreements in activity.

9/11
contrast to that, her view (which she calls perspectivalism, but is really a version of non-indexical contextualism) "does not relativize sentence truth to contexts of assessment; it relativizes only to contexts of use.So what is said on a particular occasion has an absolute truth-value with respect to the context of that occasion."(2008, p. 404).According to non-indexical contextualism, the truth-value of a particular normative judgment doesn't change every time a new assessor assesses it.In assessor-relativism it does.
Thirdly, and finally, it is worth remembering the basic features of moral subjectivism, as discussed above in the text.We are looking for the best relativistic account of moral disagreement that matches the subjectivist framework and, as we saw, subjectivism explicitly conceives of normative justification from the standpoint of the agent. 17That is, an external assessor, according to subjectivism, is expected to evaluate a given agent's values and actions from that agent's perspective.Therefore, from the perspective of moral subjectivism, there is really no room for a view like assessor-relativism.
Of course, there might be independent grounds to favour AR, that is, grounds independent of moral subjectivism.The first two arguments sketched above are meant to target these grounds, but they are far from exhausting the matter.I should add, however, that what this third and last point shows is that there is a firmly established family of views in metaethics (i.e., moral subjectivism) that is at odds with AR.It might not be a decisive consideration, but to my mind it adds pressure on proponents of the approach.Conversely, moral subjectivism is readily available to NIC 18 and that is a point in its favour. 19

Concluding remarks
The goal of this paper was to discuss whether there is and, if there is, what is the best semantic theory for moral subjectivism.The first task, then, was to clearly specify what was meant by "moral subjectivism".According to the adopted definition, it is the view that moral judgments are truth-apt, at least sometimes true, and made true by the mental attitudes of a relevant agent or set of agents.
This definition is quite broad, as it encompasses both universalistic and relativistic theories of moral normativity.Hence, the first topic of concern was whether this theoretical breadth within subjectivism posed some kind of obstacle to the project of a unified semantic theory for moral subjectivism.I argued that it doesn't.In fact, more than one semantic theory is compatible with radically different versions of subjectivism.So mere compatibility with subjectivism is not enough to settle matters in favour of any particular semantic theory.
Therefore, the remainder of the paper was dedicated to the task of comparing the relative merits and shortcomings of three different semantic theories, provided they are compatible with the original goal of coming up with a semantic approach for moral subjectivism.The comparison was conducted via the analysis of cases of moral disagreement drawn from the literature.What this study has revealed is that so-called Non-Indexical Contextualism (NIC) seems better equipped to accommodate intuitions in cases of moral disagreement.Additionally, NIC doesn't contradict a basic tenet of moral subjectivism, namely, the claim that moral truth and justification ought to be evaluated from the point of view of the agent.For these reasons, I conclude that moral subjectivists should adopt NIC as their preferred semantic theory of moral thought and talk. 17As discussed in the previous section, "justification from the standpoint of the agent" might be understood either as justification from the standpoint of the actual agent deliberating, or justification from the standpoint of some privileged agent, such as God or the ideal observer.In any of these interpretations, the point against assessor relativism holds, for an external assessor does not have the final say over the relevant agent regarding the truth or justification of this agent's thought or talk. 18And to IC, if it were not for the other problems it faces. 19The dispute between relativism and contextualism regarding perspectival claims is very much alive.In particular, data on retraction are often pointed to (e.g., in (MacFarlane, 2014)) as evidence in favour of relativism.However, see (Kneer, 2022) for an account of recent experimental data on retraction that favours a contextualist interpretation; and see (Dinges, 2022) for the claim that retraction data are not sufficient on their own to determine whether one should adopt non-indexical contextualism or assessor relativism.