Moral Judgments, 2Dism, and Attitudinal Commitments

There's an interesting new paper in Ethics, Moral Realism and Two-Dimensional Semantics by Tim Henning, which offers a lot of opportunities for disagreement. The abstract:

Moral realists can, and should, allow that the truth-conditional content of moral judgments is in part attitudinal. I develop a two-dimensional semantics that embraces attitudinal content while preserving realist convictions about the independence of moral facts from our attitudes. Relative to worlds “considered as counterfactual,” moral terms rigidly track objective, response-independent properties. But relative to different ways the actual world turns out to be, they nonrigidly track whatever properties turn out to be the objects of our relevant attitudes. This theory provides realists with a satisfactory account of Moral Twin Earth cases and an improved response to Blackburn’s supervenience argument.

There are some clever formal moves in the paper, but on a substantive level it does not strike me as very promising. Here's why...

(1) Henning motivates an "Attitudinal Semantics" for moral terms by noting that when attributing moral concepts to other speakers, "the associated attitudes have more weight than the overlap of the sets of things to which the terms are applied." That is, we're more likely to take someone to judge that something is good if they approve of it and pursue it. Henning takes this datum to motivate a view on which the reference of a moral term like 'good' is fixed by "our" practical attitudes (leaving open for now the precise scope of "our" moral community).

But I think it's much more natural for realists to instead explain the datum on the normative (rather than semantic) level. If something is good, then it's desirable. So when you attribute goodness to a state of affairs, you are rationally committing yourself to desiring it (on pain of incoherence). If someone applies the term 'X' to a bunch of things we take to be good, but they do not appear to have any desire for the things they take to be X, then either they are rationally incoherent or 'X' in their mouth does not mean the same thing as 'good' in ours. Charitable interpretation thus bars us from translating their 'X' to mean good, if it is not connected appropriately with their pro-attitudes. This does not require us to adopt a reductive "Attitudinal Semantics" (in Henning's sense). It simply follows from the normative commitments one undertakes when making certain normative judgments.

(2) Henning claims that his account accommodates the realist's intuition of metaphysical independence, that "being good does not consist in being the object of people’s approval", because on his view "there is a clear sense in which ‘is good’ expresses, and ‘goodness’ refers to, an objective property, that is, a property, the nature and existence of which do not depend on our practical attitudes."

But this feels like a cheap technical fix. (Compare my objections to Eliezer Yudkowsky's metaethics, in comments here.) Sure, the natural property ultimately denoted (on this account) is an objective property, but our practical attitudes play an essential role in the reference-fixing story, which is clearly contrary to the spirit of realism. Most realists who feel strongly about the metaphysical independence datum will presumably feel just as strongly about an analogous (meta-)semantic independence datum: that our practical attitudes do not play a central role in fixing the reference of moral terms.

We might also insist upon a strengthened version of Henning's "Actual Independence", which we may call the datum of Fundamental Fallibility: that "the extension of the term ‘good’ may turn out to differ from the set of things toward which the relevant community has the relevant favorable attitudes" even if the relevant community has full information, imaginative capacities, etc. (If our actual moral community stably takes pain and suffering to be ends in themselves, then it should follow -- on any "realist" view worthy of the name -- that our actual moral community is just plain wrong.) Attitudinal Semantics straightforwardly violates this strengthened datum.

(3) Henning's view is that the property of goodness varies depending on which world is actual. If our moral community is actually utilitarian, then what's right (in all possible worlds, considered counterfactually) is maximizing happiness. But if we consider as actual a world full of deontologists, it follows that rightness is instead the property of satisfying various duties (or whatever).

But this seems crazy. Just as it's a datum that moral terms have an invariant secondary intension (if maximizing happiness is what's right, then it's necessarily right: right in all counterfactual worlds), so it is a datum that they are invariant in their primary intension: If maximizing happiness is what's right, then it's a priori right (if knowable at all): right in all worlds considered as actual. Absent any defense, the contrary result seems on the face of it to disqualify Henning's view from being a plausible metaethical contender.

(4) Henning claims that his view can respond to Horgan & Timmons' "Moral Twin Earth" objection to metaethical naturalism. Suppose our moral community is utilitarian, and then imagine a possible "Twin Earth" much like ours but where people's psychologies are slightly different (e.g. less prone to sympathy, more prone to guilt) such that they are largely deontologists instead. Most naturalist views imply that 'right' in our mouth refers to the property of maximizing happiness, whereas 'right' in the mouths of our Twins instead refers to the property of satisfying various duties. But this violates the semantic intuition that we are all talking about the same thing, namely moral rightness, and turns our substantive moral disagreement into a merely verbal dispute.

Henning responds that "it is reasonable for us to interpret Twin Earthlings as being engaged in the same kind of pursuit as we are, [so] we can interpret both our term and theirs as terms that pick out the best candidate to satisfy this common pursuit."

But fully-informed agents could not reasonably interpret their Twins as engaged in the same pursuit. Suppose we know full well that as a matter of empirical fact our community is unanimously pursuing aggregate happiness, whereas their community is largely indifferent to aggregate happiness and instead takes conformity with the Ten Commandments as their ultimate goal. Our respective moral terms might (on Henning's view) share the same "sense" or primary intension, applying in each centered world considered as actual to the property that is pursued by the moral community at the specified "center". But this is just like the commonality in sense between our term 'water' and Putnam's XYZ-referring twin's use of 'twater'. In the relevant sense, we are talking about a different property from our XYZ-twins, and so it is in the moral case (assuming, for reductio, that moral terms function semantically like natural kind terms). So Henning remains just as vulnerable as other metaethical naturalists to the Moral Twin Earth objection.