Saturday, June 25, 2016

The 2-D Argument Against Metaethical Naturalism

A few years back I noted that 2-D semantics provides a straightforward refutation of synthetic metaethical naturalism (SEN):  SEN implies that moral terms differ in their primary and secondary intensions, this is clearly false (moral terms are "semantically neutral", or exhibit 2-D symmetry, in that their application to a world does not vary depending on whether we consider it as actual or as counterfactual), and so SEN must be false.

As I've been developing this argument in my paper 'Moral Symmetry and Two Dimensional Semantics', it occurs to me that 2-D semantics enables an even broader argument against metaethical naturalism.

To address the Open Question Argument, naturalists are committed to a divergence between the intension of 'good' (and other moral terms) and its meaning or cognitive significance.  Such divergences are commonplace so far as the secondary intension is concerned: 'water' and 'H2O' have different meanings, despite picking out just the same stuff across all possible worlds (considered counterfactually).  The problem for the naturalist is that such divergences do not seem justified when it comes to primary intensions. What a term picks out across worlds considered as actual seems very closely connected to the cognitive significance or sense of a term.

(For example, if we use 'the watery stuff' as shorthand for whatever fills the functional role of water, and which thus picks out H2O in our world but XYZ in Twin Earth, then note that 'water' and 'the watery stuff' have the same primary intensions and cognitive significance, despite differing in their secondary intensions.  "Water is the watery stuff" is cognitively trivial, whereas "Water is H2O" -- relating terms with distinct primary intensions -- is informative.)

The difficulty now for the naturalist is that there is no way to flesh out the primary intension of moral terms using purely naturalistic items in a way that does justice to the cognitive significance of moral terms.  If you have 'good' super-rigidly (in both primary and secondary intensions) pick out (say) happiness, then you seem committed to holding that 'good' means the same (or at least has much the same cognitive significance) as 'happiness', when intuitively they don't even seem to be in the same ballpark.

Note that non-naturalists face no such problem.  They can hold that 'good' super-rigidly picks out the sui generis property of goodness, which in turn supervenes on the natural things that are good -- a class of items that can only be identified through substantive moral insight, and not mere conceptual competence. Different moral communities may thus share this concept, picking out the same property of moral goodness, even as they disagree in their (implicit) theorizing about which natural items possess the moral property.

Any objections?


  1. Hi Richard
    It's an intriguing argument, but I'm not sure it succeeds for a number of reasons, including the question of what it means for a property to be non-natural. But I'll leave those issues aside, and briefly raise two concerns:

    1. Consider the terms "the first ordinal that is in a bijection with the continuum", and "the first ordinal that is in a bijection with the power set of the set of natural numbers". They pick the same stuff across worlds considered as actual, but they don't seem to mean the same (unless you don't count math? Please let me know if that's so). It might be argued they do mean the same, but it doesn't seem like that to me. And the fact that there is a bijection between the set of real numbers and the power set of the set of natural numbers isn't difficult to grasp, but it's not trivial, either. Moreover, we can use a far more complicated example if you like.

    2. Partners in innocence?
    Instead of "the watery stuff", we may consider "the reddy stuff", as shorthand for whatever fills the functional role of red (or red stuff), and "the morally wrong stuff", as whatever fills the functional role of moral wrongness or morally wrong stuff).
    If those descriptions aren't clear enough in this context, I'd suggest we stipulate it's whatever stuff our cognitive architecture is actually tracking when we normally make redness/moral wrongness judgments, but there are objections to that sort of account, so I prefer to leave the matter open as possible. In general, I'm just trying to compare the hypothesis that there is an analytic functional reduction of "morally wrong" vs. one of "red".
    Granted, it might be argued that we can conceive as actual that we're all mistaken about moral wrongness but not about redness, but I don't know whether that is true.
    Tentatively, I posted the following very brief example:

    I don't know whether the example works (there might be some problem with the description; maybe I need to consider it in greater detail), but it seems to me something along those lines might do the trick if that particular example doesn't.
    Granted, it might be argued that the functional reduction of color is simply more complicated than some of us previously thought, but for that matter, it might be that there is an extremely complicated reduction of immorality, and if there is such a reduction, we shouldn't expect that it would be found for a very long time if ever, though perhaps an AI might do it (though it would have to study working human brains for that, I think).

    1. Hi Angra, thanks for this!

      (1) Math is a tricky case, and I think the naturalist is in trouble if they have only mathematical examples of divergence between primary intensions and cognitive significance. They wouldn't seem sufficient reason to accept such a divergence in the non-mathematical case of moral-natural identities, if no other naturalistic identities (such as familiar scientific reductions) exhibit this behaviour.

      Having said that, the math case is philosophically super-interesting in its own right, so it's worth trying to figure out what to say about it. I agree the terms don't mean the same. So we should want there to be some difference in their associated semantic values (esp. the primary intension). One promising avenue here would be to say that what these terms "pick out", in the first instance, is one or other mathematical property, and it then turns out (as a matter of substantive mathematical fact) that the same number is indirectly picked out as having both these properties. This is roughly analogous to how I understand the moral case, picking out first a sui generis moral property, and then (only indirectly) the things that possess the moral property. We shouldn't skip the first stage and identify the primary intension of the term directly with the things possessing the property, precisely because that fails to track meaning / cognitive significance in cases where two distinct properties are a priori co-extensive.

      (Some philosophers model properties as sets of objects -- bigness is the set of all big things, etc. -- which works fine until you come up against these cases of co-extensive properties. So I do think we ultimately need to reject that conception of properties, at least when we're no longer talking about straightforwardly natural/empirical properties.)

    2. (2) Distinguish two senses of 'red': (i) the subjective appearance of phenomenal redness [qualia/sensations], (ii) the secondary property we attribute to objects in the world, i.e. as apt to cause red sensations in ordinary observers and viewing conditions, or whatever.

      Now, red-(i) is clearly not a functional role term, and likewise, I think, for various moral concepts. To see this, note that for any transparent description using just scientific functional concepts, it remains an open question whether you're having a sensation of redness or whether it is good. This isn't the case for water. Given the description of H2O as the clear drinkable liquid around here [... flesh out further details ...] it is no longer an open question whether water is H2O, for water just is whatever fills the described functional role.

    3. Thank you for your thoughtful and helpful replies too.

      Nice points about the math case.

      Regarding the color case, I was thinking of the ordinary sense of "red" (or the most common, if there is more than one).
      I don't think that's the subjective appearance of phenomenal redness to each of us. After all, in the ordinary sense, each of us can have a faulty color vision, be (or possibly become) color blind, etc. Also, when identifications of redness are suggested in terms of (say) dispositions to cause some perception in ordinary human observers, etc., they're not meant to capture what looks red to you, or to me, or to any specific person. Else, they would all be trivially false (since each single person could get sick, generally have their eyes modified so that they make mistaken color judgments, etc.).

      However, the example I construct seems (to me) to show that (probably) there is a possible world (as actual) in which some stuff causes red sensations in ordinary human observers in ordinary conditions, etc., but it's not red, and our ordinary judgments that it is red are false (though we would not be mistaken about what we perceive: we would still perceive it as red, but it wouldn't be so in our ordinary sense of the word). If that is so, one might mirror your argument against moral naturalism as follows:

      If redness (not the perception, but in the ordinary sense) were the property of being apt to cause red sensations in ordinary human observers and viewing conditions (or whatever), then the terms "red", and "apt to cause..etc", would pick the same stuff across worlds even when considered as actual. But they don't. So, that entails redness is not such property (and the same for any other property described in non-color terms).
      On the other hand, if redness is the property of being apt to cause red sensations in ordinary human observers and viewing conditions (or whatever) despite the fact that the respective terms do not pick the same stuff across worlds when considered as actual, then why wouldn't moral wrongness also be the property of being being apt to cause blame attributions in ordinary judges (or whatever; probably something more complicated)?

      Granted, you might say that in the example I give, some ordinary non-human observers would have it right all the time, but assuming we can characterize ordinary non-human observers, it doesn't seem more likely to me one come up with a world (as actual) in which all ordinary observers are mistaken about, say, the moral wrongness of a behavior, than one in which they're mistaken about the redness of some stuff, in the ordinary senses of both words (I'm not sure whether this is related to our different views on issues like Moral Twin Earth, which looks to me like people talking past each other (probably, though it depends on the specific description of the scenario); but maybe you have a counterexample to that scenario as well?)).

    4. Hmm, not sure if I'm following here. You suggest that conceivably (i.e. in some world considered as actual) objectual redness may diverge from the disposition to cause red sensations. I worry that you may just be introducing a third sense of 'red' here, but I'll put such worries aside and suppose you're right. Doesn't this just suggest that objectual redness is not "semantically neutral", or 2-D symmetric, in the way that terms like 'watery stuff' and 'good' are? After all, on your view we could no longer tell, of a world considered counterfactually, which things are red in it, until we first learned which world is actual (and so whether objectual redness is actually a dispositional property or an intrinsic property of objects). It would, in that case, behave more like the kind term 'water' than the neutral term 'watery stuff'. Just as we can imagine worlds where the watery stuff is not water, so in your world you suggest that the "reddish stuff" (what objectual redness turns out to be on the hypothesis that that world is actual) is not really (the actual property of) red.

      That's fine. I'm not committed to the view that objectual redness is semantically neutral. I am committed to the claim that moral terms are. Do you think that a similar story can be given whereby whether something is really good depends on whether we consider the world as actual or as counterfactual? Note that I think it's very easy to imagine a world (considered either as actual or as counterfactual) in which ordinary observers are mistaken about morality. Section 2 of my linked paper discusses just that: an inhumane society where sadistic cruelty towards non-human animals is broadly approved of. On the hypothesis that our society (including ourselves) turns out to be sadistic in this way, still such sadism is wrong.

    5. Richard,

      If your worry is correct and I'm introducing a third sense of "red", then I concede the objection fails just on account of that. The hypothetical scenario I constructed is meant to defeat a claim that there is an analytic functional reduction of "red" in terms of dispositions to cause certain perceptions on ordinary humans (or ordinary humans and some aliens, etc.). Assuming that the challenge succeeds, then the point would be that if your argument against the metaethical naturalist succeeds, a similar argument seems to succeed against the color naturalist.

      Regarding your reply, maybe there's been a misunderstanding, but I don't think this (i.e., if my scenario shows what I think it probably shows) implies that "red" behaves more like "water" than "morally wrong", for the following reason: if you're right about it (and I think you probably are), we can't conceive as actual a world in which we're wrong about water. But we can conceive of a world in which we're wrong about red (as actual or as counterfactual), just as we can do with regard to moral wrongness.

      As to your point that it's very easy to imagine a world (considered as actual, not as counterfactual) in which ordinary observers are mistaken about moral wrongness, I agree, but I'm uncertain of whether we can imagine a world as actual in which ordinary observers are mistaken about moral wrongness and would remain mistaken even after after (epistemic) rationally ideal reflection on the matter. That might or might not help the naturalist (maybe it depends on whether the naturalist can later reduce epistemic rationality), but in any event, that's not the "partners in innocence" objection I'm advancing.

      Granted, you might reject my assessment in the color error scenario on the basis of different intuitions, or else also object to color naturalism, but I think the second option would intuitively weaken your argument.

    6. "if your argument against the metaethical naturalist succeeds, a similar argument seems to succeed against the color naturalist"

      Can you flesh out how that would go? Recall my (most general) argument against the metaethical naturalist rests on the worry that "there is no way to flesh out the primary intension of moral terms using purely naturalistic items in a way that does justice to the cognitive significance of moral terms." You have suggested that perhaps the primary intension of (objectual) colour terms doesn't just straightforwardly track the dispositional property. So there's not an analytic functional reduction of 'red' solely in terms of dispositions. But presumably there is then some other naturalistic primary intension for these colour terms (using 'naturalistic' broadly to allow reference to qualia, as e.g. the dispositional account does). This broader account might pick out the particle-qualia-colours in worlds where there are such things, and the dispositional property in worlds like ours, etc. We can thus analytically reduce objectual colour to whatever actually fills this disjunctive "reddish stuff" role, just as water is whatever actually fills the "watery stuff" role. (It doesn't matter whether the agents in the world are able to correctly identify what fills the reddish role -- they might be systematically ignorant of key facts about their world. What matters is just that when given full information about the world we can, through an exercise of mere linguistic competence, identify what redness is on the assumption that the world in question is actual.)

      Does that help? I worry that we may be talking past each other here. (In particular, I'm not sure why you think I'm committed to there being an analytic reduction of redness specifically in terms of dispositions in order for colour naturalism to be true. It could be that the nature of objectual redness depends on which world is actual. But I don't think that the nature of morality depends on which world is actual.)


  2. I think probably we have very different intuitions on the matter, so probably you won't find it at all convincing, but the parallel I'm thinking would be as follows:
    Your argument states: "there is no way to flesh out the primary intension of moral terms using purely naturalistic items in a way that does justice to the cognitive significance of moral terms."
    Parallel: "There is no way to flesh out the primary intension of color terms using purely naturalistic items in a way that does justice to the cognitive significance of color terms."
    Color naturalist: "That's not true. We can analytically reduce objectual colour to whatever actually fills this disjunctive "reddish stuff" role, just as water is whatever actually fills the "watery stuff" role, even if it's complicated, and we don't have an account yet."
    Moral naturalist (parallel): "Similarly, we can analytically reduce objectual moral wrongness to whatever actually fills this disjunctive "wrong stuff" role, just as water is whatever actually fills the "watery stuff" role, even if it's complicated, and we don't have an account yet", or alternatively "We don't know what the role property is, and it depends on features of the world not accessible by conceptual analysis." (I think Boyd holds something along those lines).

    What I don't see is why the moral naturalist is in bigger trouble than the analytical color naturalist. You say that there is (presumably) "some other naturalistic primary intension for these colour terms". Maybe there is (I have worries about vagueness, but that's not the issue I was raising here). But if there is a reduction, whatever it is, we don't have it (i.e., we can't actually flesh it out right now, and I don't think that's going to change any time soon).
    But the (analytical) moral naturalist may well (and some will) also say there is a complicated reduction we don't yet know (or even if we won't be able to do so in the future).

    Granted, you might argue that that's implausible. But I don't know whether that's so, and more to the point, I don't see why the analytical naturalist would find it problematic (or why I should conclude she'd be in trouble). Granted, also, happiness isn't even in the ballpark. But if instead of happiness, the naturalists suggests something like "whatever fills the right role in a complex manner, which might or might not be fully accessible to conceptual analysis, and if it is, it might be so complicated that we probably won't know it for a very long time, if ever.", I'm not sure why that would fail.

    1. Thanks for fleshing out the parallel, that's helpful. I think the difference is that we can grasp the broad contours of a naturalistic primary intension for colour terms (as per my disjunctive suggestion in the previous comment), with no sense that anything is "left out". There's nothing more to being the colour of an object than picking out *this* sort of thing if this sort of world is actual, or *that* sort of thing if that sort of world is actual, etc. We can make a clear start on this project, which is why I've every confidence that upon further reflection we could flesh out the "etc." as needed without any great problems.

      Moral terms seem very different in this respect. Here identifying the primary intension with naturalistic items in various worlds doesn't even seem to be a start on the right track.

      Put another way: if we bracket quirky worlds involving particles-with-qualia, etc., then it seems clear enough that for the remainder, to speak of the colour of an object just is to speak of its disposition to cause coloured sensations. The concepts share the same cognitive significance at least within this restricted class of worlds. But you can't do this for moral terms. Even bracketing various quirky worlds, 'good' does not mean anything like 'happiness', or 'whatever we approve of', or any other naturalistic candidate. So we have no reason to believe that this naturalistic metaethical project is anything other than a philosophical dead-end.

    2. Thanks for the explanation.

      Personally, I'm not so confident we can (or at least I; maybe you and other people can) bracket the right worlds and get a correct (i.e., exceptionless) reduction of "red". Maybe we can, but I think it would be difficult to do.
      But assuming we (or some of us) can, in any case one could expect color to be much easier than moral terms. What I'm trying to get at is that even in a case you could expect to be much easier, it is very hard.

      On the issue of "good", I think "happiness" or "whatever we approve of", etc., have clear counterexamples, and clearly start on the wrong track, and I actually don't think there are serious and so precise naturalistic candidates yet (other than, perhaps, "whatever fills the role"). But I wouldn't expect for there to be such candidates, either, so I wouldn't find the lack of good candidates surprising. It would be extremely complex, and "good" is particularly vague. On that note, I don't even think "good" and "bad" in "a morally good person" and "a morally bad person" means the same as "good" and "bad" in (say) "pain is bad", "happiness is good".

      That said, when it comes to the less broad moral term "morally good person", if I had to suggest the contour of a broad candidate, it would be an extremely complex conjunction of disjunctions of dispositions to act in such-and-such manner and/or to feel in such-in-such way, etc., when contemplating, watching, etc., certain events, scenarios, etc. Clearly, it doesn't mean the same as "kind", though kindness might be involved in some of the extremely complex thing (an alternative candidate might be something in terms of extremely complex rules, though that's more likely for "morally wrong", or some term like that).

      On that note, it's not at all clear to me that (at least bracketing worlds no less weird than particles with qualia), something like "I know Jack is very intelligent and that he slowly tortures a randomly picked child to death every single day purely for the pleasure it gives him, of his own free will (no threats, no one messing with his brain, etc.), but is Jack a morally good person?" is a semantically open question. I wonder if linguistic competence in the use of moral terms even help you recognize that "morally good" doesn't apply to that in terms of conditions of application of the term, just what is linguistic competence in the case of the term "morally good person", and other moral terms?

    3. Some people may use 'morally good' as a thick concept (like 'kind' or 'honourable') that builds in certain descriptive constraints. But then it's a semantically open question whether the descriptive components of such thick concepts really deserve the positive normative valence we assume they do. So put such concepts aside: if you think its definitional that 'morality' is about helping people, then that's not the most fundamental normative concept. The most fundamental or thin normative concepts are pure valence: they concern what ought to be done or is worth pursuing, etc. I take it to be obvious that someone could be a competent language speaker and yet have such different values from us that he thought the torture of children was worth pursuing, and that people who aimed at this worthy end thereby had admirable character.

      That sort of positive normative evaluation is what I have in mind by 'good'. Good ends are worth pursuing, and good characters are admirable or worthy of emulation.

      Some things can be very complex and yet obvious non-starters for meaning the same as something else. The gap between 'is' and 'ought' is so wide that no amount of mere 'is', however complicated an arrangement it might be, could ever add up to meaning the same thing as 'ought'.

    4. While it's obvious to me that a competent speaker of moral language may believe that the torture of children is sometimes worth pursuing, even obligatory, etc. (some religious books may contain examples), it's not at all clear to me that someone who sincerely claims that the torture of children purely for pleasure is morally obligatory, (morally) worth pursuing, or something like that, is using moral language competently in that context (barring, perhaps, really weird bracketed worlds, though I'm also not sure that's needed at all), even if the person is otherwise a competent English speaker (by the way, there are otherwise competent English speakers who believe that nothing is red, even if something causes the right perceptions on ordinary observers in the actual world, and the actual world is not a bracketed weird world, etc.)

      That said, even if they might be using the language competently, it may well be that the connection is not direct, but requires some deduction that I'm doing unconsciously. For example, a competent user of mathematical terms may believe that the first ordinal in a bijection with the continuum and the first ordinal in a bijection with the power set of the set of natural numbers are different ordinals, because she made a reasoning mistake. Unconscious deductive reasoning is also very common.

      So, while it's not at all clear to me that something like "I know that Jack is a person who tortures children purely for pleasure every day, but are his actions morally praiseworthy, or morally obligatory?" is semantically open, it might be that it's semantically open, but not open after both meaning and some deductive reasoning that we usually do unconsciously are taken into consideration, and that is giving me the impression it might not be open.
      It might be that it's also open in that sense, but none of those questions are obvious to me.

      Regarding "is" and "ought", I think the question "I know it is morally wrong for any person to torture children purely for pleasure, but is it the case that all persons morally ought not to torture children purely for pleasure?" is semantically closed, so the issues seem to be whether expressions like "is morally wrong" are descriptive, and if so, whether what they describe can also be properly described in terms not usually considered moral terms.

      I don't know the answers to those. I do think if I say that a person is morally good, or morally bad, I'm describing her mind/character. Am I not making a description of less stable mental states if I say she's behaving immorally/morally wrong?

      As before, I don't know. Intuitively, I get the impression I probably am. Granted, it may be obvious to you that I'm not, and so it's in particular obvious to you that moral descriptivism is false. But it's not obvious to me.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.