2. Robust and formal normativity: a distinction

An increasingly recognized distinction within the metaethical literature is that between robust, or authoritative, normativity and merely formal normativity.Footnote ³ Formal normativity involves the existence of some standards, rules, or conditions, such as the rules of chess, etiquette, or legal procedures. Formal normativity is cheap in the sense that, for all it says, we may have no practice independent reason to care about its guidelines. If we’re not interested in following the rules of chess, for example, then it doesn’t matter if we move our pawns backwards. The mere existence of a standard is not in itself authoritatively binding.

Within a formally normative activity, we can make mistakes. But what is crucial about formal normativity is that such activities exhibit normative symmetry amongst each other. Though chess players and players of some nearly identical game, ‘schmess’, may run into practical difficulties when they play each other, there shouldn’t be any serious concerns about who is playing the ‘right’ game and who is failing. It is in an important sense arbitrary whether people decide to play chess or schmess. There is nothing intrinsically authoritative about formally normative properties.

Robust, or authoritative normativity, on the other hand, involves those normative facts or properties that ‘really tell you what to do’, that have ‘normative force’, are ‘significant’.Footnote ⁴ If someone claimed that there was no sense obeying the rules of morality when we could just as easily obey the rules of ‘schmorality’, we would see her as making some kind of serious mistake, or at least as engaging in a non-terminological disagreement. Morality is, for the non-naturalist, intrinsically binding in a way that merely formally normative activities are not. There is what we might call a normative asymmetry between morality and schmorality. This normative asymmetry illustrates that morality exhibits something more than merely formally normative activities: It is robustly normative.Footnote ⁵

We can make use of the robust/formal normativity distinction with respect to different kinds of entities. Robustly normative properties are those properties (such as being-morally-wrong) which are intrinsically binding and authoritative in the way that merely formally normative properties (such as being-a-legal-chess-move) are not. Robustly normative concepts are those concepts that are either intrinsically binding or those that refer to robustly normative properties.Footnote ⁶ In short, we might say that

In what follows, I won’t assume a particular view of which non-natural entity is fundamental.

3. Robust and Formal normativity: a lesson

I now illustrate the significance of the distinction for non-naturalists attempts to meet Non-Accidentality. Suppose that the following two claims are true:

Metaphysical Non-Naturalism is a core commitment of non-naturalism. Furthermore, setting perhaps some unorthodox views aside, most everyone accepts Formal Realism. So Metaphysical Non-Naturalism and Formal Realism are relatively safe assumptions to make at present.

As it's been traditionally understood, the epistemological objection to non-naturalism requires that the non-naturalist explain how the following can be true:

If non-naturalists can explain how Non-Accidentality is true, they have made important epistemological progress. However, the robust/formal normativity distinction can help to illustrate how Non-Accidentality must connect the non-accidentally true normative beliefs with the right normative content. Even if Metaphysical Non-Naturalism and Non-Accidentality are both established, it remains an open question whether our paradigmatically normative beliefs contain robustly normative contents. In order to vindicate an anti-skeptical non-naturalism, our normative beliefs must pick out robustly normative propositions. An epistemological story that only shows us how we can have non-accidentally true beliefs of formally normative propositions would show that Non-Accidentality is true, but it would not vindicate non-naturalist moral knowledge. This illustrates that Non-Accidentality isn’t the full story.

Instead, in order to defuse epistemological objections, non-naturalists must also meet:

Explaining how some of our first order normative beliefs could be non-accidentally true involves illustrating some positive epistemic relation between our normative beliefs and some stance-independent facts.Footnote ⁸ But for all that's been said, Non-Accidentality can be met even while a subject's normative beliefs contain merely formally normative concepts or properties. Non-accidentally true normative beliefs are a necessary component of a non-skeptical non-naturalist epistemology, but they are not sufficient.

Non-Accidentality and Content Success are distinct conditions on an adequate moral epistemology. But it is important to keep in mind that the non-naturalist must explain how both can be met with respect to many of the very same beliefs. Showing how some paradigmatically normative beliefs of type A are non-accidentally true and showing how some paradigmatically normative beliefs of type B are of robustly normative propositions is not enough. It must further be shown that there is some overlap between A and B. In other words:

Content Success, and its relationship to Non-Accidentality, have been overlooked.Footnote ⁹ Both are crucial to a proper understanding of how non-naturalists must respond to the epistemological objection. This interplay between Non-Accidentality and Content Success is an important one.

The robust/formal distinction, then, is important for non-naturalist epistemologists to keep in mind, regardless of their other metaphysical commitments. Content Success must be met by anyone committed to Metaphysical Non-Naturalism, a core commitment of all non-naturalists. I turn now to illustrating how one recently popular strategy for meeting Non-Accidentality, the Conceptual Competence Strategy, fails. I then discuss how the objection applies to three particular versions of the Conceptual Competence view.

4. Conceptual competence and Content Success

Appealing to the epistemic conditions on conceptual or semantic competence has been one recently popular strategy for explaining Non-Accidentality. Footnote ¹⁰ The very rough idea is this: It's overwhelmingly plausible that we are competent users of our normative language, as well as the concepts that figure in our normative beliefs. Conceptual and semantic competence entails that certain epistemic conditions are met. These epistemic conditions ensure that some of our normative beliefs are justified and non-accidentally true. Further information and reasoning can help us to gain more justified and non-accidentally true normative beliefs. Non-Accidentality is met in virtue of our conceptual and semantic competence in the normative terms. To see how this is supposed to work in practice, let's consider an example from Foot:

Hedgehog is clearly false. But, the thought is, it isn’t just false, but it couldn’t be true (barring some radical change in the laws of nature). As Foot says, ‘there is some content restriction on what can intelligibly be said to be a system of morality.’Footnote ¹² Interpreted in one way, this comes to the claim that the denial of Hedgehog is something of a conceptual truth. (I take no stand on whether Foot herself is committed to this.) And we can know this merely in virtue of our conceptual competence of WRONG.

The proponent of competence based epistemology, as I’ll call it, wants to extend the structure of Foot's case to a variety of first order normative propositions. The details, both about the extent of substantive conceptual normative truths, as well as our ease or difficulty in grasping them, vary widely between different competence based epistemologists. I will discuss whether these details make a difference for the success of my objection in the following sections. For now, I present the objection in outline, before considering its application to a few prominent competence based epistemologists.

Non-naturalists, as we’ve seen, believe that moral properties are both robustly normative and stance-independent. That means that, for all that's been said, the normative concepts that we’ve developed may not have latched onto the robustly normative properties. Unless we have some assurance that our concepts have latched onto the robustly normative, rather than some merely formally normative properties, showing that we have knowledge grounded in normative conceptual competence does not make any genuine epistemic progress.

The normative concepts we have developed are, for all that the conceptual competence theorists have said, contingent. Those in a different evolutionary, social, or cultural niche may have had importantly different normative concepts, and with them, importantly different conceptual truths that could be grasped on the basis of their own conceptual competence.Footnote ¹³ At most one of these distinct conceptual frameworks latches onto the robustly normative. And, for all that conceptual competence theorists have said, we have no reason to believe that it is us, and some reason to think that it isn’t. The three case studies also illustrate and reject a few potential ways to avoid the objection.

5. Huemer's A Priori moral epistemology

Michael Huemer has attempted to explain our epistemic access to non-natural moral facts via a general theory of a priori knowledge. Huemer identifies properties with universals.Footnote ¹⁴ Furthermore, as with most non-naturalists, he accepts that ‘the moral realm is causally inert.’Footnote ¹⁵ And though he admits that moral properties, if they exist, would be metaphysically strange, he doesn’t see this as evidence against their existence.Footnote ¹⁶ Finally, he accepts the received view that moral properties supervene on natural properties, though they do so in a non-reductive way.Footnote ¹⁷ An evaluative belief that p, on Huemer's view, has some initial positive epistemic status so long as it is grounded in an intellectual seeming that p.Footnote ¹⁸

Huemer's explanation of a priori knowledge involves four important claims:

Taken together, these claims can explain how moral beliefs meet Non-Accidentality. Here is how Huemer explains it:

Since forming an adequateFootnote ²² concept involves meeting purely intrinsic criteria, and meeting those very criteria constitutes adequately grasping a (mind-independent, causally inert) universal, Huemer has given us an explanation for how we can form non-accidentally true beliefs about universals without standing in some causal or quasi-perceptual relation to them: All that matters is whether the concepts which constitute the belief in question meet the (reflectively accessible) conditions, and we’ve got non-accidental truth.

An immediate question arises with respect to (C). Even if my concept C is consistent, clear, and determinate, that doesn’t yet explain how C is guaranteed to have a corresponding (mind-independent) universal. As it turns out, for Huemer, this can’t happen, because ‘[t]here is no possibility of one's [concept] failing to refer to anything (universals are plentiful in this sense, and their existence is necessary).’Footnote ²³ As long as our concept is adequate, we can be sure that it refers to a real mind-independent universal. Huemer's way of meeting the Non-Accidentality, then, relies on a fifth important claim, which I’ll call Plenitude:

With this explicit statement of Plenitude, we have a complete story about how a priori moral beliefs can explain Non-Accidentality.Footnote ²⁴ Since we are capable of reflecting on our moral concepts, adjusting them to become more adequate, and since we know that these moral concepts have corresponding mind-independent universals (of which our adequate concepts constitute the grasping of them), we can have non-accidental a priori beliefs about at least some of the truths in which our moral concepts feature.Footnote ²⁵

We’ve seen Huemer's general story about how Non-Accidentality will be met. With an eye to Content Success, let's turn to a particular case of how epistemic access to moral facts works on Huemer's account. Consider a straightforwardly ‘good’ case of how a particular ethical intuition can guarantee the truth of its contents. Suppose Lucy is reflecting on her concepts of BADNESS and LYING, and on that basis it seems to her that:

Furthermore, let's assume that her concepts of LYING and BADNESS are consistent, clear, and determinate, Footnote ²⁶ and that BADNESS for Lucy plays the sort of roles we traditionally associate with normativity. Then, by Plenitude, they both refer to mind-independent universals. Let's call the universals they refer to being-a-lie and badness, respectively. Since Lucy's concepts are consistent, clear, and determinate, and she formed an intuition based on reflecting on those concepts, the following is guaranteed to be true:

So we have gone from a claim about Lucy's mind-dependent concepts to a mind-independent fact about universals. And presumably, since she has met the conditions that guarantee the truth of my belief that (L*), she knows it, or at least is justified in believing it.

So far, so good. But now consider Carol. Carol is also engaged in ethical inquiry, and also reflects on her concepts of BADNESS and LYING. Carol's concepts are also clear, consistent, and determinate. But reflecting on these concepts, she has the intuition that:

By Plenitude, then, for similar reasons, the following is guaranteed to be true:

Since (-L*) is guaranteed to be true, and Carol has formed her belief in a way that guarantees it to be true, she also knows it, or at least is justified in believing it.

It should be obvious at this point that something has gone wrong. (L*) and (-L*) are incompatible with each other, so they can’t both be ‘guaranteed’ to be true – at least one of them has to be false. And I certainly don’t mean to suggest that Huemer is unwittingly committed to contradictory ethical claims both being true. Rather, there is a subtle but informative mistake in the reasoning given above. Lucy and Carol both grasp their respective concepts clearly, consistently, and determinately. This means that, so long as they are being careful, they won’t misapply those concepts. Their concepts have a clear and determinate extension, and they will both be able to classify things as picked out by their concepts or not. But, by Plenitude, so long as their concepts have clear and determinate extensions, there will be properties that their concepts track. So they do both form true beliefs about the extension of their concepts. However, as the difference between (L) and (-L) shows, the extension of their concepts of LYING and BADNESS differ. And this doesn’t show us that one of them is mistaken – after all, their respective concepts are all fully adequate – rather, it shows us that they have different concepts altogether.

The right way to classify the case, then, is to distinguish Lucy's concepts from Carol's concepts. Call Lucy's concept BADNESS_L and Carol's concept BADNESS_C.Footnote ²⁷ With this distinction in mind, we can more precisely characterize their respective ethical intuitions:

Given Plenitude, both Lucy's and Carol's concepts correspond to mind-independent properties. And since their concepts are adequately grasped, they will both have knowledge of the corresponding mind-independent facts. However, since their concepts differ, the corresponding properties will differ as well. The knowledge that Lucy has gained from her ethical intuition is not best characterized as (L*), but as:

Furthermore, the knowledge that Carol has gained from her ethical intuition is not best characterized as (-L*), but as:

Once we get clear about what Lucy and Carol believe based on their intuitions, then, we see that they do not contradict each other after all. This is analogous to the merely formally normative chess case discussed above. The schmess players truly believe that knights can move diagonally, while chess players do not. But this is because the schmess players have beliefs about being-forbidden-in-schmess, while chess players have beliefs about being-forbidden-in-chess. If the schmess players and the chess players were in an argument about this, they would be merely talking past each other.Footnote ²⁸

We can multiply cases like Lucy and Carol's indefinitely. There are many different fully adequate possible ‘moral’ concepts, each of which differs about the extension of ‘bad’. Huemer can’t deny the multiplicity of adequate ‘moral’ concepts, since adequacy is merely about internal consistency. These concepts are all compatible with one another, so long as we are clear about whether we are talking about bad ₁, bad ₂, etc. However, unlike in the chess case, or the case of any formally normative concept, there is a further question in the moral case: Which concepts adequately characterize the robustly normative properties? By Plenitude, any adequate moral concept will correspond to some mind-independent property or other. Huemer has shown that (e.g.) Lucy's beliefs meet Non-Accidentality. Her belief (L_L*) is a non-accidentally true first order normative belief. But he has provided no reason to accept that Lucy's normative concepts actually refer to the robustly normative properties, and thus no reason to accept that her normative beliefs were of the robustly normative facts. In other words, though Huemer has shown how Non-Accidentality can be met, he has not shown how Content Success can be explained, and so has not provided a wholly adequate non-naturalist epistemology. To paraphrase Street, so long as it hasn’t been shown that there is a relation between the moral concepts we happen to have and the normative facts, the appeal to conceptual competency offers no way, in the absence of an incredible coincidence, of showing how our moral beliefs could meet Content Success.Footnote ²⁹

However, there is a disanalogy here between the moral case and the chess/schmess case. Moral properties, unlike the rules of chess, are robustly normative–indeed, that is where our problem arises. But why couldn’t Lucy pack into her concept that it refers to the robustly normative badness, if it refers at all? This would ensure that she can meet Content Success–it's a condition on her concept that it refers to the robustly normative badness, after all. And since this addition doesn’t render her concept incompatible with the other conditions on conceptual adequacy, it looks like the explanation of Non-Accidentality is still met as well.Footnote ³⁰ Problem solved?

Not quite. As we saw above, Huemer's strategy for meeting Non-Accidentality essentially relies on Plenitude. An adequate concept can’t fail to refer to a property, because there is a property for any extension that a given concept might fix. Any adequate concept is sure to hit a target property, because there are an abundance of targets. If we give up Plenitude, there becomes a non-trivial chance that an adequate concept will fail to target any property, and so won’t ensure non-accidentally true beliefs, at least about mind-independent properties.

In order for Huemer's strategy to work, then, the proponent needs to ensure that the conditions she places on the adequacy of a concept leave open the abundance of possible referents. Consider a toy example. Polly the Pythagorean has the concept TEN_P. Her concept is extremely similar to our concept TEN. She takes it to refer to the successor of nine, the thing that results from five added to five, etc. In fact, Polly and us agree about all of the first-order mathematical facts. However, Polly has a further view about the nature of the referent of TEN_P. Unlike our own concept, it is internal to Polly's concept that TEN_P refers to a sacred and holy property, if it refers at all. If she were to become convinced that Pythagoreanism is false, she would see her TEN concept as defective.

There is certainly no in principle problem with the existence of concepts like TEN_P. However, insofar as there are such concepts, they are incompatible with Plenitude. Suppose Polly's TEN_p concept is completely adequate in Huemer's sense. This would not be enough to guarantee true beliefs, since it must also be the case that the concept's extension latches onto a property that is holy and sacred. And nothing about the internal conditions on Polly's concept can guarantee that such a property exists. Plenitude says that there is a property for every extension. But it doesn’t say that there is a holy and sacred property for every extension. So Polly's concept will not ensure Content Success, if holiness is a non-negotiable internal feature of her TEN_p concept.

For similar reasons, moral concepts like Lucy's BADNESS cannot contain, as an internal condition, that it refers to robustly normative badness if it refers at all. If this can be an internal condition on Lucy's concept, it can on Carol's as well. But, either one of them is mistaken about the extension of their concept, or one of their concepts fails to refer (because, for example, there is no robustly normative property of badness that applies to lying). If the former, Non-Accidentality is not met.Footnote ³¹ If the latter, Content Success is not met. So adding this further condition on our normative concepts will not solve the problem after all.

Notice that the success of the objection does not rest on the possibility of divergence between two agents such as Lucy and Carol. The case of Lucy and Carol is merely meant to help illustrate the problem. The problem is that a wholly adequate non-naturalist epistemology requires more than the mere having of normative beliefs that track some facts non-accidentally. It must be that those beliefs track the robustly normative facts non-accidentally. Even convergence amongst agents would not be enough to explain how our normative beliefs are responsive to the robustly normative facts. For all that's been said, and for all we know, our beliefs in such a situation would be converging on some merely formally normative facts.

6. Schroeter & Schroeter's metasemantic solution

Schroeter and Schroeter (Reference Schroeter and Schroeter2017) suggest that the moral realist can meet Non-Accidentality by appeal to Chalmersian metasemantics. Although their view is strictly incompatible with non-naturalism, Footnote ³² a slight variation on the view to be discussed below allows the non-naturalist to make use of the same general strategy. Even though Schroeter & Schroeter are not themselves non-naturalists, their view is important to discuss because it is perhaps the most developed version of what I’m calling the conceptual competence strategy.

The crucial principle that they appeal to in their epistemological story involves an idea originally found in Chalmers and Jackson (Reference Chalmers and Jackson2001):

The empirical information must be given in an ‘ideal base-level descriptive vocabulary’, rather than in natural language, on pain of circularity. An ideal base-level descriptive vocabulary is ‘an exhaustive microphysical description of the actual world, together with a ‘that's all’ clause and an ‘I am here’ clause.’Footnote ³⁴ I’ll grant for the purposes of this paper that such a vocabulary is in principle possible.Footnote ³⁵ The basic idea behind (IA) is that the reference of a subject's concepts ‘respect[s] the subject's own ideal, fully informed verdicts about the reference of her words and concepts.’Footnote ³⁶

In order to resolve epistemological objections to non-naturalism, IA only needs to be true of normative terms. Suppose, as Schroeter & Schroeter argue, it is. Recall that Non-Accidentality requires some explanation as to how our actual normative beliefs can track the moral properties, given that those properties are causally inert. IA tells us that a subject's moral beliefs under ideal conditions are guaranteed to be true, given that those beliefs themselves determine the referent properties of the moral concepts that figure in them. So our ideal selves’ moral beliefs can straightforwardly explain Non-Accidentality. However, this also means that some subset of our actual moral beliefs are non-accidentally true as well. This is because

Some of our actual moral beliefs must survive the idealization process, lest those beliefs not be our ideal moral beliefs. This means that some of our actual moral beliefs are non-accidentally guaranteed to be true, though we don’t know which ones.Footnote ³⁸ We also have reason to believe, given IA, that further reflection and information will help weed out more and more false beliefs. Our giving beliefs upon receiving more information is very good evidence that they wouldn’t survive ideal reflection, and thus are not true. IA helps establish that some of our actual moral beliefs must be true, and also that further reflection and information is likely to result in more true moral beliefs.Footnote ³⁹ These results are enough to meet any reasonable epistemological demand on the non-naturalist, and thus the non-naturalist who endorses IA can meet Non-Accidentality.

Schroeter & Schroeter see IA as best fitting with a naturalist realism. This is because they endorse the identity of necessarily coextensive properties.Footnote ⁴⁰ Given that there will be some set of natural property instances across all possible worlds that is extensionally equivalent to the normative properties, there will be a straightforward identification between any normative property and a natural property. Of course, as Schroeter & Schroeter themselves admit, non-naturalists (as well as others) have rejected this claim, lest their views collapse into naturalism.Footnote ⁴¹ Schroeter & Schroeter are worried that rejecting the identity of extensionally equivalent properties will cause problems for IA, since the procedure of IA only seems to generate an extensional set of property instances across worlds. So it is too coarse grained of a procedure to fit with the rejection of an extensional equivalence theory of property identity.

I think the non-naturalist that is otherwise attracted to something like IA should push back on this claim. For the non-naturalist does think there is an important and graspable difference between the normative properties and the natural properties that are extensionally equivalent–namely, the normative properties exhibit robust normativity, whereas the natural properties do not (and perhaps cannot). So long as competent users of normative terms take robustness to be crucial to what it takes to be, for example, wrong, they can make a more fine-grained distinction between normative and natural properties than extensional equivalence will allow. There may be reasons to think competent users of normative terms don’t have a grip on robust normativity in this way. But it seems plausible that non-naturalists would think that they do, and Schroeter & Schroeter have not given an argument to think that this is not so. The non-naturalist should be able to help themselves to IA after all, contra Schroeter & Schroeter.

With all that in mind, let's return to the distinction between formal and robust normativity. Schroeter & Schroeter's view faces a dilemma, each horn of which is incompatible with the truth of either Non-Accidentality or Content Success. It must turn out that (a) at least some of our substantive first-order normative beliefs are true, and (b) the properties picked out by our moral terms are the robustly normative properties. We’ve seen above how IA could ensure that (a) is met: Since our ideal beliefs are a result of some function on our actual beliefs coupled with empirical information, there is a quasi-constitutive connection between our first order normative beliefs and the properties that figure in them.

Suppose for the sake of argument that (L) is one of Lucy's beliefs that would survive idealization. Call the property that figures in Lucy's ‘bad’ thoughts badness _L. Recall that Carol believes that:

And suppose for the sake of argument that (-L) is one of Carol's beliefs that would survive her own idealization.Footnote ⁴² Call the property that figures in Carol's ‘bad’ thoughts badness _C.

As with Huemer, Schroeter & Schroeter ascribe to something like Plenitude, so there is no worry about Lucy or Carol's terms failing to refer.Footnote ⁴³ However, the same problem arises for Schroeter & Schroeter's view that arose for Huemer's view. They have given a metasemantic story which can explain how to avoid a skeptical moral epistemology, but it makes meeting Content Success extremely difficult. Lucy and Carol both meet Non-Accidentality. Some set of Lucy's actual normative beliefs S figures in (partially) determining her idealized normative beliefs. By IA, the members of S are non-accidentally true. However, if the normative beliefs that make up S don’t match the first order robust normative truths, Lucy's normative beliefs won’t meet Content Success. While some of Lucy's normative beliefs would be non-accidentally true, it would be extremely unlikely for those beliefs to be of the robustly normative facts. But Carol's epistemic situation is identical to Lucy’s. Schroeter & Schroeter have provided an epistemology that can explain Non-Accidentality but not Content Success.

7. Cuneo and Shafer-Landau on ‘moral fixed points’

In an ambitious recent paper, Terence Cuneo and Russ Shafer-Landau's crucial claim is that ‘there is a battery of substantive moral propositions & that are also nonnaturalistic conceptual truths.’Footnote ⁵⁰ They go on to argue that if this claim is accepted, it does a wide-ranging amount of metaphysical and epistemological work for the non-naturalist. (Here I am only worried about the alleged epistemological work that can be done.) For example, the following claims are, they argue, excellent candidates for being non-naturalistic conceptual truths:

First-order non-naturalistic conceptual truths such as these they call the moral fixed points. Any normative system which fails to endorse the moral fixed points would thereby not count as morality. Cuneo & Shafer-Landau argue that the moral concepts provide substantive constraints on moral theorizing.

A natural question arises: What reason do we have to believe that our moral concepts pick out anything stance-independent? Cuneo & Shafer-Landau's answer to this question appeals to what they call the ‘traditional view’ of concepts.Footnote ⁵² On the traditional view, concepts have three important features. First, concepts are the constituents of propositions.Footnote ⁵³ Second, concepts are referential devices that ‘enable thinkers to refer to things such as objects and properties.’Footnote ⁵⁴ Finally, concepts are ‘abstract, sharable, mind-independent ways of thinking about objects and their properties. As such, they are very much objective, ‘out there’ sorts of things, extra-mental items whose existence does not depend on our employing them in thought or language.’Footnote ⁵⁵ As mind-independent entities, concepts have essences which underlie conceptual truths such as the moral fixed points.Footnote ⁵⁶ This last feature of the traditional view is crucial for answering the above question. Our question was how individual moral representations can pick out anything mind-independent. The answer, once the traditional view is granted, is simple: Concepts themselves are mind-independent, so once we have a competent grasp of one, we already have a competent grasp of the essence of a mind independent entity.

We are now in a position to see how Cuneo & Shafer-Landau can use the traditional view along with the moral fixed points to explain Non-Accidentality.Footnote ⁵⁷ Take some moral fixed point m. Now suppose I am conceptually competent with respect to the concepts (normative and otherwise) that figure in m. Suppose, furthermore, that upon reflection I accept m. My belief will be non-accidentally true: I formed it in light of my conceptual competence, and my conceptual competence ensures that, if I don’t make some mistake, my belief is true. So my belief that m is not just true, but is true non-accidentally. And such an explanation will extend to the beliefs of all conceptually competent agents when their beliefs are of moral fixed points.

Cuneo & Shafer-Landau's view explains Non-Accidentality, at least with respect to the moral fixed points. But can these beliefs simultaneously explain Content Success? As we’ve seen, on Cuneo & Shafer-Landau's view, the truth-makers for a moral fixed point are just the concepts that make it up. No further ‘worldly fact’ is needed to ground its truth. This is what constitutes the fact that moral fixed points are conceptual truths. But neither are conceptual truths merely analytic truths. Analytic truths are sentences which are true in virtue of the meaning of the words that compose them. Analyticity, as Cuneo & Shafer-Landau understand it, is a linguistic, not metaphysical, phenomenon. Conceptual truths, on the other hand, are metaphysically robust in that they involve mind-independent essences. Since concepts have essences, and, as in the moral fixed points, concepts’ essences bear relations to each other, it seems like a genuine possibility that moral concepts could themselves be robustly normative, on Cuneo & Shafer-Landau's view. If this were correct, then no further relation between the moral concepts and any non-natural properties would need to hold for robust normativity to get in the picture. The moral concepts would just have robust normativity built into their essences.

Grant that the traditional view of concepts is correct and that moral fixed points are true in virtue of the essences of the concepts that constitute them. Furthermore, grant that there are concepts which have robust normativity built into their conceptual essence. These three assumptions can clearly support an explanation of Non-Accidentality, but what about Content Success? It might seem as though Content Success can be explained, once we’ve granted that robust normativity is built into a concept's essence. By now the problem is a familiar one. Showing that we have substantive conceptual moral knowledge and that some concepts are robustly normative is not enough: We need to show that our moral concepts are the robustly normative ones. For again, there could be any number of alternative sets of broadly speaking normative concepts which deny some or all of the moral fixed points while affirming others. There are, for example, the schmoral concepts, and with them, the schmoral fixed points, which are conceptual truths grounded in the essences of the schmoral concepts.Footnote ⁵⁸ We could enumerate such non-moral but broadly speaking normative sets of concepts, fixed points, and essences indefinitely. What we need is some reason to believe that our moral concepts, as opposed to any of these other sets, are the robustly normative ones. This is what explaining Content Success requires, and it is difficult to see how it would fall out of Cuneo & Shafer-Landau's view.

Interestingly, Cuneo & Shafer-Landau acknowledge a related point, asking ‘Why think that we have reason to pledge our allegiance to this normative system, rather than another – call it schmorality – that fails to incorporate & the moral fixed points?’Footnote ⁵⁹ (I take it that they mean ‘reason’ here in the robustly normative sense.) In response to this question, they concede:

In itself, this is a fair enough concession: A single paper can’t defend every controversial piece of non-naturalist metaphysics. However, this concession has a crucial epistemological upshot. We have no explanation – by Cuneo & Shafer-Landau's own lights – as to why we should think our moral concepts are robustly normative. There are any number of alternative systems of normative concepts, each from our epistemic standpoint equally likely to be the robustly normative concepts. For all that's been shown so far, we have more reason than not, probabilistically speaking, to think that our moral concepts are not robustly normative. This would mean that without further argument, Cuneo & Shafer-Landau have shown that (many of) our beliefs in the moral fixed points are non-accidentally true, but of no more normative interest than our beliefs about the rules guiding how one can move one's pawn in chess.Footnote ⁶¹

Funding

This work was supported by the Israel Science Foundation [grant number 1972/17].