Parfit’s and Scanlon’s Non-Metaphysical Moral Realism as Alethic Pluralism

Thomas Scanlon and Derek Parfit have recently defended a meta-ethical view that is supposed to satisfy realistic intuitions about morality, without the metaphysical implications that many find hard to accept in other realist views. Both philosophers argue that there are domains of discourse in which true judgements do not have ontological implications. In the scientific domain true judgements are about reality and these judgements have ontological implications. By contrast, Parfit maintains that in the normative domain these truths have Bno positive ontological implications^ and are not Babout metaphysical reality^ (2011 Vol. II, p. 479 & 747). And for Scanlon, normative truths Bneed no natural or special metaphysical reality in order to have the significance that we commonly grant them^ (2014, 52). Although it is not evident that Scanlon and Parfit hold the same view, many (including Scanlon himself) have taken them to do so. Therefore, in what follows I will assume that there is one core idea that the quotes from the previous paragraph try to capture. I take the projects which develop this core idea to be an attempt to show that positing non-natural entities is not necessary to satisfy realistic intuitions. What distinguishes Scanlon and Parfit’s approach from other realistic normative theories is that both maintain that normative entities exist in a way that is different from (some) nonnormative entities. Moreover, they think that the way normative entities exist helps to answer the metaphysical worries that normally plague non-naturalism. I agree, however, with Cowie (2014) andWedgwood (2016) that Parfit and Scanlon fail to develop their view sufficiently. So it is the purpose of this article to investigate in what way their core idea can be developed while at the same time satisfying realistic intuitions. To do this, I will proceed as follows. I will first specify what realistic intuitions are and what it takes to satisfy them. I will do this by looking at the debate about mind-independence between meta-ethical realists and anti-realists (section 2). Ethic Theory Moral Prac (2017) 20:751–761 DOI 10.1007/s10677-017-9801-6

I will propose that a view satisfies realistic intuitions about morality if it gives a better account of mind-independence than anti-realists do. Then I will present Scanlon's response to the usual metaphysical objections to non-naturalism. I will call the basic idea developed here domain pluralism (DP). In section 4 I will develop an account that grounds DP in alethic pluralism: the position that there is more than one truth property. I will call the resulting view Alethic Domain Pluralism (ADP). In section 5 I will show that this view fails to satisfy realistic intuitions about morality. This shows that on a plausible and initially promising reading of what it takes for a normative entity to exist in a way that is different from non-normative entities, Parfit's and Scanlon's non-metaphysical moral realism fails to be more realistic than contemporary versions of anti-realism.

Realistic Intuitions
Although intuitions about the status of moral judgements vary greatly, many philosophers agree that moral truth is objective, by which they mean that it is mind-independent. Just as it is not our way of thinking about peaches that causes them to fall from the tree when ripe, it is also not our way of thinking about morality that makes it the case that some moral judgements are true. This is one of the realistic intuitions about morality that is not only held by moral realists, but also by many anti-realists. These anti-realists subsequently try to show that even a metaethical view that denies that moral entities are a fundamental part of reality can account for the intuition. In this section I will look at an argument, proposed by the moral anti-realist Simon Blackburn, which purports to show that his meta-ethical theory can also account for mindindependence. 2 I will then show why moral realists typically think that this explanation does not give a proper account of mind-independence. I will use this discussion to formulate a criterion for satisfying realistic intuitions: an account satisfies realistic intuitions if it gives a better explanation of mind-independence than the one anti-realists typically give.
In his Spreading the Word, Blackburn defends the view that the meaning of moral sentences can be cashed out in terms of desire-like attitudes such as approval and disapproval. On his account, the sentence Bkicking dogs is morally wrong^expresses disapproval of kicking dogs. As Blackburn recognizes, one might think that such an account is committed to the minddependence of moral truths, but he sets out to show that this is not the case. He maintains that whether an account is committed to mind-dependence can be spelled out in terms of commitment to conditionals such as the following: KICK Bif we had different sentiments, it would be right to kick dogs^ (1984,. To resist (KICK), he argues that Bas soon as one uses a sentence whose simple assertion expresses an attitude, one is in the business of discussing or voicing ethical opinion^ (1993,173). Interpreted as a moral assertion, (KICK) expresses the attitude that one approves of letting information about people's sentiments dictate one's attitude towards kicking dogs. So on Blackburn's account, the anti-realist can resist conditionals such as (KICK) because from her perspective kicking dogs is repugnant and from this perspective letting information about sentiments dictate how to feel about kicking dogs is just as repugnant. So Blackburn's argument is based on two premises: (1) that resisting conditionals such as (KICK) is enough to satisfy mind-independence and (2) that these conditionals are true or false in virtue of one's actual moral attitudes or perspective. We can therefore characterise Blackburn's notion of mind-independence as follows: Perspectival mind-independence The actual perspective of the assessor (one's moral attitudes) determines the correctness of moral standards, even in evaluations of counterfactual situations.
In response to this move by Blackburn, critics usually try to show that even if Blackburn's strategy is successful, the anti-realist cannot satisfy another form of mind-independence. The critic argues that on the anti-realist's account, someone with very different moral attitudes could faultlessly judge that kicking dogs is permissible. In other words, even if from my actual perspective kicking dogs is repugnant, there could be another person who approves of kicking dogs and from her actual perspective, this would be correct. And this would still be a kind of mind-dependence according to the critic. Let us define this notion of mind-independence as follows: Non-perspectival mind-independence Normative standards are correct or incorrect, if they are, from all possible perspectives.
The anti-realist can of course claim that the negation of non-perspectival mind-independence (in other words, non-perspectival mind-dependence) is also a moral claim and resist it for this reason. The realist insists, however, that there is a theoretical or descriptive reading of this claim and that the anti-realist cannot satisfy it on this non-normative reading. Some antirealists deny that there is such a reading, but another option for the anti-realist is to accept the availability of this reading while denying that it is the preferred reading in everyday contexts. For example, Horgan and Timmons (2006, 89) claim that in detached contexts we cannot assign a truth value to moral claims such as non-perspectival mind-independence. This is where I take it that the debate has reached a standoff. I shall not further engage in this debate. What I will do is explore realist alternatives for the philosopher who shares the intuition that some versions of moral anti-realism fail because they do not satisfy nonperspectival mind-independence but who also wants to avoid the (allegedly) dubious metaphysical entities that most versions of moral realism are committed to.

Domain Pluralism
Let us start out with a characterization of the meta-ethical position that we can find in Scanlon (2014) andParfit (2011). In Being Realistic about Reasons, Thomas Scanlon defends a realistic cognitivism about (non-instrumental) reasons. The view is cognitivist since it holds that Bclaims about reasons for action can be correct or incorrect^(2014, 2). Moreover, Scanlon holds that reasons are not reducible to non-normative truths and cannot be explained in terms of more fundamental notions such as rationality or rational agency, at least not insofar as these latter notions are not themselves claims about reasons.
The view that there are irreducible normative truths faces several worries, which are typically thought to be metaphysical, motivational and epistemological in character: we can ask whether the existence of irreducible normative truths is compatible with a scientific view of the world, whether we can be motivated by such irreducible normative truths, and how we can come to know such truths. To address these worries, Scanlon proposes the following: I believe that the way of thinking about these matters that makes most sense is a view that does not privilege science but takes as basic a range of domains, including mathematics, science, and moral and practical reasoning. It holds that statements within all of these domains are capable of truth and falsity, and that the truth values of statements about one domain, insofar as they do not conflict with statements of some other domain, are properly settled by the standards of the domain that they are about. Mathematical questions, including questions about the existence of numbers and sets are settled by mathematical reasoning, scientific questions, including questions about the existence of bosons, by scientific reasoning, normative questions by normative reasoning, and so on (2014,19).
In this passage Scanlon argues that accepting a scientific view of the natural world does not commit one to the view that truth-apt statements can only be about the natural world. In addition to the scientific domain there is a range of other basic domains, such as the mathematical and the moral domain, within which statements can also be true of false. Moreover, the truth of these statements is relative to the standards of the specific domain to which these statements belong. Let us call this basic idea developed above, domain pluralism: Domain Pluralism There is a plurality of domains, including mathematics, science, the moral and the normative domain. If we say that an entity exists or that a statement is true or false this is relative to one of these domains.
In order to develop this idea we need an explanation of what it means for an entity to exist in a domain-relative way. I will present one plausible construal in the following section.

Alethic Domain Pluralism
In the previous section I have suggested that the best way to understand Scanlon's meta-ethical position is as the thesis that existence is domain-relative, i.e., always relative to a specific domain. The upshot of this thesis is that it allows us to say that the domain-relative existence of normative entities is not ontologically problematic, since it does not commit us to the position that normative entities exist in the same way as entities in the natural world. In this section I develop this idea by showing that domain-relative existence can be grounded in the more fundamental notion of domain-relative truth. Since I will raise an objection to this view in the next section, it is not my intention to fully defend this position. What I will do, however, is make clear how such a view can be motivated.
To see how domain-relative existence can be grounded in domain-relative truth, we have to interpret Scanlon's claim that Btruth values of statements about one domain, [...] are properly settled by the standards of the domain that they are about^ (2014,19). Standards that are appropriate in the scientific domain, such as correspondence, might not be appropriate when applied to the mathematical or moral domain. So possession of a specific property, such as coherence with other statements, might be necessary and sufficient for truth in one domain, while this property might be neither necessary nor sufficient in another domain. On this interpretation we say that different properties play the truth-role in different domains of discourse. Or, in other words, domain pluralism amounts to a form of alethic pluralism, the position that there is more than one truth property. Alethic pluralism is a position in the debate about truth that has benefited from the pioneering work of Wright (1992) and has recently been developed by Lynch (2004Lynch ( , 2006Lynch ( , 2009; Pedersen and Wright (2013).
There are several ways in which alethic pluralism has been spelled out in the literature. One way to think of alethic pluralism is to take the term Btrue^to be ambiguous between different domains of discourse. 3 So this means that Btrue^in the scientific domain would mean something different from what it means in the normative domain. However, many people are suspicious about positing lexical ambiguity without independent evidence. Another way to think of alethic pluralism is to maintain that although Btrue^is univocal, it picks out different properties in different domains. Crispin Wright's reductive pluralism (1992) is an example of such a position. He maintains that the meaning of Btrue^is determined by a set of platitudes about truth. 4 One forceful objection to such a view is that it cannot explain what property is preserved in valid mixed inferences. Consider an instance of modus ponens where the antecedent of the conditional is normative and the consequent is a scientific claim. For example, cruel cats are hungry, this cat is cruel, therefore this cat is hungry. 5 If we adopt a simple version of alethic pluralism, we can no longer say that the argument is valid, since there is not one truth-property preserved. As a response to this objection, several other accounts have been proposed. But, the details of the correct version of alethic pluralism do not bear on the main thread of this paper. For this reason, I will assume in the rest of this paper one of the most influential, recent versions of alethic pluralism: Michael Lynch's Alethic Functionalism. According to this view, truth is a functional kind and different properties can play the truth-role in different domains. To avoid the mixed inference objection, the alethic functionalist holds that truth should be identified with truth-role itself, not with the property that plays this role. 6 My argument will, however, not rely on the details of this specific account. 7 The main idea behind Alethic Functionalism is that different domains of discourse have different properties playing the truth-role. So although it might be the case that in the scientific domain a statement is true if it corresponds to mind-independent facts, other domains, such as mathematics and the normative domain have different properties playing the truth-role. One influential way of spelling out another notion of truth is the pragmatist notion of truth that was originally defended by Charles Sanders Peirce. In its original formulation it says that B[i]nquiry properly carried on will reach some definite and fixed result or approximate indefinitely toward that limit^(1931, 1.485) and B[t]he opinion which is fated to be ultimately agreed to by all who investigate, is what we mean by the truth^ (1931, 5.407). 3 This view has been defended by Tarski (1944) and Kölbel (2008). 4 These platitudes include at least the following: (1) to assert is to present as true, (2) truth is absolute and does not admit of degrees and (3) truth-apt contents have truth-apt negations and can be embedded within conditionals, conjunctions, etc. 5 The example is from Tappolet (2000, 383). 6 For different versions of Lynch's functionalist view, compare (2001Lynch's functionalist view, compare ( , 2004 and (2009,74). 7 For the argument I am presenting in this paper, it does not matter that truth is a specific role that is played by different properties in different domains. What matters is that in different domains there are different properties that determine which sentences are true. For this reason the argument generalises to all other recent versions of alethic pluralism.
In recent years this formulation has been updated and defended by truth pluralists such as Crispin Wright and Michael Lynch. The most recent version is superwarrant and is defined as follows: Superwarrant P is superwarranted just when believing P is warranted at some stage of inquiry and would remain warranted without defeat at every successive stage of inquiry.
As Wright has argued, this property satisfies most platitudes that we usually associate with the truth-predicate. A belief can, for example, be warranted without being superwarranted, since having warrant for a belief is consistent with it being defeated at a later stage of inquiry. Also, superwarrant does not admit of degrees and the negation of a content that is capable of being superwarranted is itself also capable of being superwarranted. This pragmatist theory does not commit one to a particular view on warrant, but it is natural to combine this theory of truth with a coherentist theory of warrant. Scanlon, for example, argues extensively that the method of reflective equilibrium fits best with the subject matter of morality (and mathematics).
If we take superwarrant to be the property that plays the truth-role in the normative domain and correspondence to play the same role in the scientific domain, we can motivate a form of domain pluralism. To see how, consider two very different criteria for existence. 8 On the one hand we have the Eleatic principle, endorsed by David Armstrong, Brian Ellis and Hartry Field. 9 According to this principle we can only allow entities into our ontology if they have causal powers. On the other hand, we have a principle that is ascribed to Gottlob Frege, where we take truth to be constitutively prior to reference. 10 According to this Neo-Fregean Principle, we allow entities into our ontology if they appear as the referent of a singular term in a true sentence. 11 I take Scanlon to endorse a slightly broader version of this principle in the following passage 12 : as long as [a] way of talking [is] well defined, internally coherent, and [does] not have any presuppositions or implications that might conflict with those of other domains, such as science, [we would] by accepting these statements [...] be committed to the existence of things quantified over in the existential statements counted as true in this way of talking (2014,27).
We can unpack this idea by using the following procedure: we first determine the true statements within a domain and we then determine what the ontological commitments of this domain are. To determine which statements are domain-relatively true, we have to check whether some specific conditions are met; these conditions are internal (well defined and 8 See also Cotnoir and Edwards (2015). 9 See also Colyvan (1998) for a critical assessment of the principle. 10 See Eklund (2006, 98) 11 The Neo-Fregean priority principle adhered to in this passage should not be confused with the reconceptualization thesis that is also associated with Neo-Fregeanism. See also Eklund (2006, 98-99). 12 The principle Scanlon refers to implies that entities that are existentially quantified over should be admitted into our ontology. The Neo-Fregean principle only holds for entities that are named by singular terms. Given existential generalization, the Neo-Fregean principle implies Scanlon's principle. Moreover, as existential quantification plays an important role in Scanlon 's principle, it resembles Quine's criterion for ontological commitment. However, on most interpretations, Quine does not endorse the thesis that truth is constitutively prior to ontology, so these theses should be distinguished.
internally coherent) as well as external (no conflicting presuppositions or implications). From these true statements we can derive a set of objects that are quantified over in true existential statements, and, because of this, we can say that these objects exist in a domain-relative way. To give an example, we can truly say BThere is an even prime number^, or more simply, B2 is a number^. Because these sentences satisfy the criteria mentioned above, we can say that the number 2 exists. So this is a reversal of the platonic idea that the sentence B2 is a number^is true in virtue of the existence of a number. I will call the thesis that truth is (at least in the normative domain) constitutively prior to ontology the truth priority thesis.
So if the above is correct, we have two ways of thinking about the relation between truth and ontology. In the scientific domain existence is determined by the Eleatic Principle: we start out with a realm of causally efficacious objects and we can say that certain statements are true in virtue of the existence of these objects. We can thus say that in the scientific domain ontology grounds truth; or ontology is more fundamental than truth. On the other hand, in the normative and mathematical domains we start out with true sentences. Among these true sentences are existential claims. We can now say that normative and mathematical existence is grounded in the truth of these existential claims. Therefore, in the normative and mathematical domains truth is more fundamental than ontology. Given that existence in the scientific domain grounds truth, and truth grounds existence in the mathematical and normative domains, it follows that we are talking about different ways of existence. So, given the alethic pluralist development of Scanlon's domain pluralism, there are (at least) two ways of being: a more realistic way in the scientific domain and a weaker way in the normative and mathematical domains.

Disagreement
In this final section I raise an objection to Alethic Domain Pluralism. This objection is based on the possibility of moral disagreement. Above, we saw that according to the alethic domain pluralist, a normative judgement is true if and only if it is superwarranted. This is an epistemic account of truth: an account that understands truth in terms of epistemic notions such as knowledge, belief or justification. Epistemic accounts of truth typically rely on the method of reflective equilibrium as a theory of justification. And Scanlon and Parfit are no exception to this rule. 13 A well-known problem with the method of reflective equilibrium is that it is not obvious that different individuals with different starting points would converge on a unique reflective equilibrium. 14 And this point is clearly related to my discussion in section 2, where I argued that in order to satisfy realistic intuitions a theory should satisfy the following desideratum: Non-perspectival mind-independence Normative standards are correct or incorrect, if they are, from all possible perspectives.
A theory satisfies non-perspectival mind-independence if it can show that all starting points that are possible converge on a unique reflective equilibrium. Scanlon acknowledges this requirement, and admits that if a divergence of equilibria would take place on a massive scale, we would face two options: it would force us to (1) suspend judgement on the issues that we disagree about or (2) accept a form of pluralism regarding this set of judgements (2014,(79)(80). Scanlon seems quite confident, however, that in practice the reflective equilibria will tend to converge.
One way of ensuring convergence is to rule out inappropriate starting points. This strategy would require that we have a way of distinguishing between appropriate and inappropriate starting points. However, whether a starting point is appropriate or not is a normative question. Consequently, we have two options concerning the truth of the claim that a starting point is appropriate: either it is itself determined by the process of reflective equilibrium, or it is not. If it is not, then there must be an independent account of normative truth which distinguishes the appropriate from the inappropriate starting points. It is then difficult to see why we would need the process of reflective equilibrium. If the truth of the claim is determined by the process of reflective equilibrium, then we can ask the same question again about the appropriateness of the starting points of this process. So we would end up in a vicious infinite regress.
An alternative strategy is to argue that there is already a rough overlap in the initial starting points for the reflective equilibrium. A defender of this position could back up such a claim by citing Sharon Street's observation that there are Bdeep and striking patterns, across both time and cultures, in many of the most basic evaluative judgments that human beings tend to make ( 2006,115). So even if we had to admit that not all moral judgements have determinate answers, there is a core of universally held beliefs that are part of all actual reflective equilibria reached. This is what Scanlon calls a Brealistic cognitivism about reasons: a view that is cognitivist in holding that claims about reasons for action can be correct or incorrect, but realistic also in recognizing that there may be limits to the range of cases in which such claims have determinate truth values^(2014, 2). So although there are some statements in the normative domain that do not have determinate truth values, there is at least a solid core of normative truths.
The story would then be as follows: given a rough overlap in initial starting points, individuals will converge on a set of basic normative beliefs. It might be the case that on details we reach different conclusions, but there would at least be a solid core of shared normative beliefs. The convergence demand then seems to be satisfied. But is this enough to satisfy realistic intuitions? As we can see from the definition above, non-perspectival mind-independence is a claim about all possible perspectives. So the claim is ambiguous between different modalities. Perhaps it is, given our evolutionary history, impossible that we would have moral starting points that would lead to different reflective equilibria. This would mean that it is historically impossible that we would have cognitive and conative states that would make different normative standards correct. If this is the right kind of modality, non-perspectival mind-independence is satisfied. However, although it is (arguably) historically impossible that we had evolved otherwise, we can still question whether it is also nomologically (according to the laws of nature) or metaphysically impossible that we had different moral starting points. So the question becomes what kind of possibility is needed to satisfy realistic intuitions. I will argue that historical necessity is not enough to satisfy these intuitions.
To see why this is the case, let us consider the following thought experiment. Consider a normative clash between us and a different society. Suppose further that this society has normative beliefs that we find perverse; for them, sacrificing an innocent child for the greater good is what ought to be done. We then want to say that this society has mistaken normative beliefs. If we denied this, we would be accepting perspectival mind-independence. However, given the assumptions laid down in the current section, we know that if this other society has different normative beliefs, this must be the result of a mistake in their normative deliberation: since they started out with roughly the same moral starting points as we did, they should reach roughly the same reflective equilibrium.
So far so good. Now suppose that we find out that this society had a slightly different evolutionary history; we discover that the members of this society are not members of our species. Perhaps historical research shows that this society is a not yet discovered branch of the Cro-Magnon-Man. Apart from this historical fact which instilled different evaluative tendencies, this society functions much like other societies found on earth. Given the possibility of the scenario sketched above, with two communities that use normative words with the same normative role but different extensions, Eklund (2012) has distinguished two responses someone could give. Firstly, we could say that there is a further question as to which society really is right. Secondly, if there is no such further question, we could claim that we are right and the other society is wrong, but we would have to admit that they could say the same about us and be as justified as we are. 15 Let us consider the first response, where there is a further question as to which society gets things ultimately right. A problem with this option is that the further question seems inexpressible. Let us take Bought*^to be the word from the other community with the same normative role as our ought and X to be the action that we find perverse, then we ought not to do X, but we ought* to do X. And of course, we really ought not to do X, but we also really ought* to do X. All these claims might be true on this account, even objectively so, but we have the feeling that at least some community is making a huge mistake. 16 Moreover, if this further question were not inexpressible, it would be completely unclear what would determine the correct answer. Both parties' judgements are, after all, true.
So let us now consider the second option. On this option the evolutionary history of an agent determines which actions are right for her. So relative to this history and the evolved moral starting points, different standards are right for the two different societies. In other words, if we had other cognitive or conative states, that perspective would make different standards correct. So on this second horn, ADP turns out to be unable to satisfy nonperspectival mind-independence, the view that we set out to avoid. We can therefore conclude that temporal or historical necessity is not enough to satisfy realistic intuitions.
Another way of making the same point is to notice that if mere historical necessity were enough to satisfy non-perspectival mind-independence, the anti-realist position from section 2 can also satisfy the realistic intuitions about mind-independence. Remember that the argument above relied on the contentious premise that because of our shared evolutionary history, there is a huge overlap in our moral starting points. And this gives us non-perspectival mind-independence that understands the modality involved as historical necessity. But if we accept this, then the anti-realist can also argue for non-perspectival mind-independence. Given a robust way of forming more complicated moral beliefs from basic beliefs, the anti-realist can also maintain that it is historically impossible that we had different cognitive or conative states which would make the application of different normative standards correct. But then the alethic pluralist loses her advantage over the anti-realist.
Moreover, ADP also faces another problem. Suppose that contrary to what I have just argued, ADP could be developed in such a way that the starting points of the reflective equilibrium were necessarily the starting points that we currently have, in a stronger sense of 'necessity' than historical. For example, suppose that these starting points can be derived from 15 Actually, Eklund distinguishes a third option in which there is no further question but we just deny that the situation between us and the other society is symmetrical. Since Eklund also takes this to a very implausible alternative and it is very hard to see what could motivate such complacency I ignore this option is this paper. 16 See Eklund (2012, 144). rational agency and that we are necessarily rational agents (although Scanlon and Parfit do not show that this is in fact the case). In this scenario every possible alternative community would have the same starting points. Would this satisfy realistic intuitions? To see that it would not, we can imagine that one of these alternative communities has a slightly different property that plays the truth-role in the normative domain. Let us say that for them, superwarrant* 17 plays the truth-role in the normative domain. And now it could be the case that Bwe ought not to do X^is superwarranted, while Bwe ought to do X^is superwarranted*.
If we assume that this scenario is possible, then here again, the two options that we saw earlier emerge. We can question whether superwarrant is the correct predicate for playing the truth-role in the normative domain, or whether superwarrant* ought to fulfil this role. Of course, it is superwarranted that superwarrant ought to fulfil this role, but it is similarly superwarranted* that superwarrant* ought to fulfil this role. So the inexpressibility objection rears its head again. If, on the other hand, we insist that there is no further question whether superwarrant of superwarrant* is the correct predicate for the normative domain, the result is, again, a failure to satisfy non-perspectival mind-independence.
Above we have seen that a view that applies the method of reflective equilibrium to our own moral starting points has undesirable consequences. To remedy this problem we could modify the view in question so that the moral starting points of all people should be taken into account. We could formulate such a view as follows: moral judgements are true if and only if they are superwarranted and a judgement is warranted if and only if it was formed by applying the method of reflective equilibrium to the moral starting points of all people.
This view escapes the critique of the previous section, since by definition we cannot have different stable equilibria with contradicting judgements. However, it has other unacceptable consequences. A first problem is whether we should include the moral starting points of all actual or all possible people. If we include only actual people, then removing a part of the population would be one way to achieve equilibrium. If we include all possible people, we again face the problem that we have to decide what kind of possibility is relevant here. For example, if all logically possible moral starting points would be taken into account, then it is hard to see how a reflective equilibrium could ever be reached.
Moreover, even if such a view satisfied realistic intuitions, it would clearly fail to satisfy our normative intuitions. Imagine again that we encounter a (much larger) society with perverse normative beliefs. How should we proceed? One option is to adopt, at least partially, the views of the community that we at this moment find horrible. Another option is to weigh the strength of the convictions people have, but this would again imply that the people with the strongest convictions get to dictate what is morally right or wrong. Or perhaps if the views are irreconcilable, we should suspend judgement. In all of these cases, we would end up with moral views that seem obviously wrong, merely because we encounter a group of people with a different moral outlook. Of course, we can sometimes learn from other cultures and travel broadens the mind, but morally abhorrent views should be denounced rather than adopted. For this reason I conclude that applying the method of reflective equilibrium to all moral starting points of different agents is unsuccessful as well.

Conclusion
In this paper I have discussed a plausible and initially promising way of spelling out Parfit's and Scanlon's thesis that there is a difference in ontological commitments between the normative and the scientific domain. This approach grounds domain pluralism in alethic pluralism. We have seen that it provides a plausible development of the claim that normative entities exist in a Bnon-metaphysical^or Bnon-ontological^way by showing that in the normative domain truth is more fundamental than ontology. I also argued, however, that this position fails to satisfy realistic intuitions. This result is, of course, conditional on a specific reading of what it takes for an entity to exist in a non-ontological way. So my conclusion can also be read as an invitation to develop this idea in a more persuasive way.