Sharon Street has argued that (prominent versions of) quasi-realist expressivism (“expressivism,” for short) has a knowledge problem. Others, such as Allan Gibbard and Jamie Dreier, deny this. I’m inclined to agree with Sharon on this matter, albeit for reasons somewhat different from those she offers when she states her challenge.
Stating what the alleged problem is requires a bit of set up. The first step is to get a sense of what distinguishes quasi-realist expressivism from realism. Here is one way to do so that meshes with much of what expressivists say.
Reflection on everyday moral thought and practice seems to reveal that:
(i) In moral experience, the world calls for or demands a certain kind of response. Agents often form moral judgments on the basis of such “morally rich” experience.
(ii) Moral judgments have the marks of a descriptive belief: they are classificatory, truth-evaluable, apt candidates for knowledge, and apt for inference.
(iii) Moral reasoning often generates the conviction that not any response will do and, inter alia, the conviction that we can make moral mistakes.
(iv) Moral judgments have the marks of a practical attitude: they are often directive, and motivationally efficacious.
Call these claims, which metaethical theories endeavor to accommodate and explain, core moral data. Realists endeavor to accommodate and explain claims such as (ii) by appealing to the claim that moral judgments are beliefs with moral moral representational content and they attempt to accommodate and explain claims such as (iii) by holding that there are moral facts that we can accurately (or fail to accurately) represent. Call beliefs and facts of these sorts robust moral beliefs and robust moral facts (because of the explanatory work they’re supposed to do). Expressivists attempt to accommodate and explain claims such as (ii) but not by appeal to the claim that moral judgments are beliefs with moral representational content. And they attempt to accommodate and explain (iii), but not by appealing to moral facts. Expressivists generally agree that we have moral beliefs, that they represent moral facts, and that moral facts exist. But they deny that these beliefs and facts are robust in the sense just specified; data such as (i)-(iv) get accommodated and explained by other means.
Now consider a claim such as:
(A) It’s a fact that engaging in recreational slaughter of fellow persons is wrong.
Expressivists maintain that (A) is true. But there are at least two different positions to which expressivists might commit themselves when maintaining this.
Thin expressivism embraces a metalinguistic/semantic thesis about what it is to say or think (A) is true. Roughly, the view tells us that to commit oneself to (A) is simply to commit oneself to being in an attitudinal state of a certain kind (e.g., a planning state, or a state of condemnation, or….) and whatever else that entails (where this does not include robust moral facts). Thick expressivism goes further, maintaining that to commit oneself to (A) is to commit oneself to there being moral facts (albeit not robust ones).
I worry that those writing on behalf of expressivism move between these views when replying to the sorts of concerns that Sharon raises. But I’ll touch upon that later. For now, suppose we understand expressivists to embrace Thick expressivism. As I understand it, Thick expressivism endorses both:
(B) Moral judgments often accurately represent the moral facts; and
(C) When moral judgments accurately and reliably represent the moral facts – that is, “track” the moral facts – these judgments are often cases of moral knowledge.
The view, then, accepts claims such as:
(a) Your judgment that engaging in recreational slaughter is wrong tracks the fact that engaging in recreational slaughter is wrong.
(b) Your judgment that torture is wrong tracks the fact that torture is wrong; and:
(c) Your judgment that lying simply to save face is wrong tracks the fact that lying simply to save face is wrong.
The initial worry I’d like to raise is that, if Thick expressivism were true, the truth of these propositions would be a remarkable coincidence because in no case could the moral facts which your moral judgments account track account (even in part) for the accuracy and reliability of your moral judgments. This worry, as I say, is close to but not identical with Sharon’s because it does not appeal to anything in the neighborhood of an evolutionary debunking argument. It simply appeals to (a), (b), and (c) and the claim that moral facts exist but are not robust.
To say that moral facts cannot account for the moral knowledge we have sounds strange to me: How could one know that recreational slaughter is wrong but not because recreational slaughter is wrong? But that aside, Thick expressivism seems to have two options available in response to the charge that it commits itself to a vast array of remarkable coincidences. The first is to accept some version of a “third-factor” view, according to which there is some entity that both explains why the moral facts hold and how we could track them (but does not entail that moral facts themselves explain how we track them). The second option is to embrace some version of a response-dependence version of moral facts, according to which (correct?) attitudinal states determine the moral facts that we track and that we can track these attitudes (but does not entail that moral facts themselves explain how we track them).
These responses move us beyond the initial worry because now explanations are in the offing. The further worry, though, is that (I) there is no version of a third-factor view that’ll be compatible with Thick expressivism and (II) committing itself to a response-dependent view would expose Thick expressivism to exactly the sorts of objections that realists press against other varieties of response-dependence view and that, in various places, expressivism claims to avoid altogether.
One reason why I’m inclined to agree with Sharon that there’s a problem here is that those who respond on expressivism’s behalf tend to do so by appealing to Thin expressivism. They tell us that there’s no problem because expressivism is simply telling us what it is to say or think claims such as (A) and what it would be for (A) to be “objective.” If Thin expressivism were the view under consideration, that seems like a promising response. But if Thick expressivism is the view under consideration, then appealing to metalinguistic/semantic claims regarding what it is to think and say things like (A) doesn’t address the worry raised at all. These claims could not be offered as explanations of (a)-(c). A still further and related worry is that it really seems that when expressivists present their view, they often have Thick expressivism in mind; they affirm that there are moral facts (sometimes even designating what these facts are, namely, true thoughts.) If these appearances are to be trusted, then when expressivists reply to the remarkable coincidence challenge by appealing to what Thin expressivism says, they are taking back with one hand what they give with the other.
Terence,
This is interesting. I`m curious why you think that a thick expressivist would deny that moral facts are robust and whether, if they wouldn`t, you think your reliability challenge would also hit against non-expressivist quietist-types and non-naturalists.
Two thoughts:
First, it sounds to me like you’re saying that expressivists get in trouble when they go in for “thick expressivism”, over and above “thin expressivism.” But I’d have thought part of the point of expressivism is the idea that there isn’t actually a difference there; to be committed to the existence of moral facts just is to be committed to being in attitudinal states of certain kinds. (I mean that “just is” to be read symmetrically–these are two ways of describing the same commitment.) So if you thought there was no problem for the thin expressivist, then you also shouldn’t think there’s a problem for the thick one.
Now maybe you’ll say we should just go in the other direction–there is a problem for the thick expressivist–she’s basically just a realist–so if there’s no difference between thick and thin expressivism, then the thin expressivist has a problem too. This is where my second thought comes up–it’s basically a version of Jack’s question. I think there’s a strand of expressivism–especially strong in Blackburn–that doesn’t try to distinguish itself from realism, expecially non-naturalist quietist realism. Like the quietist realist, this sort of expressivist won’t say much about the metaphysics of moral facts (except negative claims–they’re not natural facts, they’re not reducible to some other kind of facts). Unlike the quietist realist, she’ll say a lot about the philosophy of mind–here’s what it is to be in the state of mind of believing in a moral fact, and here’s how it differs from believing in a descriptive fact, etc.
If that’s right, then the expressivist shouldn’t hold that she avoids a problem that arises for realists. Instead, she should say that she’s explained why there’s no problem here for anybody, realists included. And the strategy goes, roughly, by getting you to agree that there’s no epistemic problem for thin expressivism (as you seemed on board with), and then arguing that there’s no difference between thin expressivism and thick expressivism.
Hi
interesting. A question of clarification: what’s meant by tracking? It cannot be anything of the form – had the moral facts been different, then the judgments would have been different too. Moral facts, I take it, are necessary facts that could not have been otherwise. So, is there a way to formulate reliability and tracking in a way that doesn’t make it a matter of correlation through possibilities?
Second concern. I take it that expressivists are likely to think that our talk about tracking, reliability and knowledge are through and through normative and thus they are going to give an expressivist, semantic treatment of what we mean when we make the kind of claims used in the argument. If this is right, then there is nothing stopping the expressivist accepting (a) to (c) as normative statements.
I’m inclined to agree with all three of the comments above.
With Dan, I agree that it’s hard to see how thin expressivists are supposed to be understood, since once committed to (A) it seems extremely perverse to deny that there are moral facts (or to deny that one is committed to moral facts, having confessed to a commitment to (A)). So then (as Dan says), with Jack I start to wonder what the denial of robustness is supposed to amount to, either for an emaciated expressivist or a quietist. (I do have a view about what that denial might amount to, but I don’t think yours fits with mine.)
And with Jussi, I wonder what you have in mind by ‘tracking’. My first thought is that expressivists should join (some) robust non-naturalists in insisting that we know in some other way than tracking, in moral cases; but that’s because I’m thinking of tracking as involving counterfactual dependence.
But, bottom line, I kind of agree with the spirit of the post. Expressivists do not have an explanation of the Big Coincidence, which is the fact that our fundamental moral beliefs are remarkably well correlated with the moral truth. My thought, not very well worked out, is that expressivists are apt to have a better story about why there is no such explanation than, say, a robust realist can offer.
Jamie
here’s one thought about the last comment. Would expressivists not give a moralized explanation of the Big Coincidence? Could they not, for example, use the Aristotelian story about how through appropriate upbringing, education and practice we require all kinds of ethical sensitivities that enable us to discern the morally relevant features of the situations we face and their relevance in those circumstances? This story would be couched in ethical terms but it would still be at least somewhat informative as it would explain why getting things right is not a coincidence. Furthermore, this seems like just the kind of story that would be metaethically neutral and thus available for expressivists, non-naturalists and possibly others too.
I’m inclined to agree with Jamie that expressivists might be better positioned to give a story about why there is no such explanation. I also think that they are in a better position to insist that it isn’t problematic to postulate such a coincidence. An expressivist might admit that there is a grand coincidence between their attitudes and the moral facts, but the epistemic significance of this coincidence is very different than the significance of other coincidences, and it doesn’t ultimately undermine their view.
I think we’re far too ready to apply traditional epistemological tools and arguments in evaluating expressivism without noting the significant difference in status of normative facts. If we ultimately explain belief in moral facts in terms of non-representational attitudes, then the belief that there is a grand coincidence between the moral facts and our moral attitudes isn’t itself (robustly) representational, unlike other beliefs in grand coincidences. It might instead be best understood as some kind of complex feeling. So I suspect it has a very different significance to what other things we should feel or what else we should believe than beliefs in other kinds of coincidences.
I’m sympathetic to many of the comments above, and especially to the suggestion that expressivists should not claim that they avoid epistemological problems that realists face: once we’ve started the minimalist ascent into realist territory, the convergence with quietist realism, and even with non-naturalist realism, seems inevitable to me. It’s true that both Blackburn and Gibbard have often claimed that their view does not collapse into full-blooded realism, and that this is why they are immune to reliability challenges that realists fall prey to. I just don’t see how they can make good on these claims, lest they treat realism as something akin to Dworkin’s “moral field theory”: a view according to which moral facts play a causal role in the best explanation of our beliefs. That view would clearly be in conflict with Blackburn and Gibbard’s evolutionary account of moral thought, but it is also something that most realists, including robust non-naturalist realists, reject.
Assuming then that there is no divide between expressivism and realism, is there something in expressivism that helps explain why we shouldn’t be worried about a reliability challenge like Street’s? If so, these resources would equally be available to realists. Unfortunately, I’m not sure what they are. Simply appealing to positive psychological claims about conative attitudes does not seem to address the problem: expressivists still owe us an explanation of how we came to have by-and-large correct attitudes (or by-and-large true beliefs about the moral facts, to use the realist ideology that expressivists accept). I’m tempted to say that the negative functional theses of expressivism might be of help here: say, the etiological function of moral thought is not to track moral facts, but rather to help solve coordination problems, and this is why we ought not to demand an explanation of reliability like the naturalistic tracking accounts available in other regions of discourse. But I’m not sure how dialectically effective this would be as a response to those who insist that any explanation of our reliability, no matter if naturalistic or not, should ensure the safety and sensitivity of our beliefs.
Hi Jack,
Thanks for yours. In characterizing expressivism in the way I have, I’ve attempted to state how expressivists themselves often state their position. (E.g., Blackburn when he writes that expressivism is “visibly anti-realist, for the explanations offered make no irreducible or essential appeal to the existence of moral ‘properties’ or ‘facts.'” , EQR 175, 198). I think of expressivism as committing itself to two projects – call them the accommodation and simulation projects. The accommodation project consists in accommodating/explaining the core moral data. The simulation project consists in thoroughly mimicking what realists say when they accommodate/explain the core data. However, the mimicry extends only up to that point at which explanatory claims come into the picture. So, yes, to the existence of moral facts, but no to moral facts explaining how our moral beliefs could be accurate, mistaken, and so on. If expressivism were to sign up for the explanatory claims – and, hence, commit themselves to robust moral beliefs and facts – then there would be no more mimicry. Then the view would be a version of realism.
Yeah, I think the reliability challenge does pose a challenge to quietists and certain versions of non-naturalism.
Terence,
It’s true that expressivists often describe the contrast between their view and realism in explanatory terms, but they usually take the relevant explananda to be our beliefs or linguistic practice. I don’t think they would deny that moral facts play a role in explaining how our moral beliefs could be accurate/mistaken. They usually treat any such explanation as an exercise in first-order reasoning and thus are just as likely to invoke moral facts here as realists are.
Expressivists do want to claim that moral facts play no role in psychological explanations of our beliefs, or at least in an evolutionary story about the emergence of normative thought and our evaluative tendencies. But it’s not clear to me why non-naturalist realists would disagree on this point, given that they normally hold that moral facts are not the sort of things that can play a causal role in psychological or anthropological stories.
Thanks for the interesting post. I would like to follow up on Jussi’s first comment. You write,
“The initial worry I’d like to raise is that, if Thick expressivism were true, the truth of these propositions would be a remarkable coincidence because in no case could the moral facts which your moral judgments track account (even in part) for the accuracy and reliability of your moral judgments. “
I do not know how to evaluate this worry — or an analogous one which conditionalizes on “robust realism” — without a better sense of what you (or Sharon or…) mean by “account”. As Jussi notes, it is presumably not necessary, in order to relevantly “account” for the truth of our moral beliefs, that we argue that had the explanatorily basic moral truths been different (i.e., had the conditional truths which state the conditions under which moral properties are instantiated been different), our moral beliefs would have been correspondingly different. The problem is not that such counterfactuals are unintelligible. The problem is that we cannot argue for such counterfactuals in uncontroversial cases. For instance, we cannot argue that had particles arranged computer-wise failed to compose a computer, we would have believed this. On the other hand, if “accounting” for the truth of our moral beliefs merely requires arguing that, holding the explanatorily moral truths fixed, we could not have easily had false moral beliefs, then we may well be able to account for their truth. Insofar as genealogical considerations help to show that we could not have easily had different epistemically basic moral beliefs, they may actually help us to do this. (Nor should we need to argue that had our moral beliefs been different, they would still have been true. Such a counterfactual is antithetical to realism, and, anyway, we cannot argue for it in the perceptual case.) Finally, “accounting” for the truth of our moral beliefs should not simply require showing that their contents (or truth) is implied by the best explanation of our having them. By that standard, it is trivial to account for the truth of every (true) logical belief, since every logical theorem is implied by every explanation at all.
Perhaps you think that “accounting” for the truth of our moral beliefs requires showing that their contents (or truth) is implied by the best explanation of our having them, but requires more too? Perhaps it requires showing that the contents of our moral beliefs play an “explanatory role” in the best explanation of our having them. But, then, what do you mean by “explanatory role”? You cannot mean causal role, unless you wish to rehearse Benacerraf’s argument, which just assumes that knowledge of causally inert entities is impossible. If you appeal to hyperintensional ideology like grounding, then I doubt your worry will serve Street et al’s purposes (it is hard to argue that learning that the fact that p fails to ground your belief that p undermines that belief, under a realist construal). I worry that by “explanatory role” you will just mean whatever role is necessary to account for the truth of our moral beliefs.
Hi Terence,
Thanks for this post! I’m a little concerned about your characterization of expressivists who have responded to Street. You say that they “[appeal] to metalinguistic/semantic claims regarding what it is to think and say things like (A)”.
But this is not what Blackburn says in his response to Street. In fact, he says precisely the opposite, adopting the first-order normative perspective and asking whether evolution has pushed him towards the moral truth: “I suppose that evolutionary forces or other contingencies made me such as to value happiness more than misery, and good for them, since at least in this instance they have pushed me towards a truth.” This is no metasemantic or linguistic claim. It’s “thick” expressivism in action, not thin expressivism.
It is true that an explanation for our reliability is required, but I think that Blackburn, along with naturalist-realists who have responded to Street, denies that there is any sense in asking about our reliability from a purely descriptive or “thin” perspective. I believe that he would say that since reliability is a normative notion, it can only be assessed from within a normative framework, one that takes certain truths for granted (this is just the rejection of what he calls “Cartesian realism”). Here, I confess that I don’t see why a third-factor response (or perhaps a variant of Jussi’s story) is necessarily unavailable to the expressivist, though of course not just any old story will do.
Thanks, Daniel, for your response. I think that Thin expressivism and Thick expressivism are different views and are open to different sorts of challenges. Here’s a shot at explaining why.
Let’s assume that realists can say what they mean when they claim that there are moral facts and that what they say is, by all appearances, different from what expressivists say when they say that there are moral facts. (There are certain places where Blackburn denies this, holding that realism isn’t really a distinct position at all).
I take it that all versions of expressivism commit themselves to thoroughly mimicking what realism says. That’s what I call above the simulation project. (Lots of mimicry language in Blackburn and Gibbard, for example.) It looks to me that this mimicry must have two dimensions. It must involve, first, engaging in the sorts of activities in which realists engage when realists attempt to accommodate/explain the core data. So, it must involve saying things such as “moral facts exist” when presenting their view (but without committing themselves to robust moral facts in so doing). But it must also involve offering or presenting mimics of that to which realism commits itself when it claims that moral facts exist. If mimicry didn’t have this second dimension, it would be possible that what expressivism commits itself to when it says that moral facts exist in no interesting sense resembles or approximates what realism commits itself to when it claims that moral facts exist.
What Thin expressivism commits itself to when it claims that moral facts exist are attitudinal states of certain kinds (and whatever else these commitments might imply that doesn’t involve the existence of robust moral facts). The worry is that attitudinal states are not good candidates _at all_ for being “fact-mimics” (whatever that might come to). Since attitudinal states are pretty bad candidates for being moral facts (as realism thinks of them), it’s best to interpret Thin expressivism as simply offering metalinguistic/semantic claims about what it is to say or think that moral facts exist (which do not imply that those who issue them commit themselves to moral facts).
Thick expressivism, as I understand it, explicitly commits itself to there being fact-mimics, these incorporating moral properties, which Blackburn says are mere “shadows” or “abstractions” from moral predicates. This view, I think, has a shot at thoroughly mimicking what realism says. However, as I’ve characterized things, if Thick expressivism denies that moral beliefs and facts are robust, then this is not a version of realism. (Again, I take myself to be following expressivists in this respect.)
So, the versions of expressivism I’m interested in exploring are ones that do try to distinguish themselves from realism. And they differ from one another because one version commits itself to fact-mimics (Thick expressivism), while the other (Thin expressivism) doesn’t.
It may be that there are yet other versions of expressivism that don’t try to distinguish themselves from certain broadly non-naturalist versions of realism.
Justin, I had trouble understanding what you meant by the following-
“As Jussi notes, it is presumably not necessary, in order to relevantly “account” for the truth of our moral beliefs, that we argue that had the explanatorily basic moral truths been different (i.e., had the conditional truths which state the conditions under which moral properties are instantiated been different), our moral beliefs would have been correspondingly different. The problem is not that such counterfactuals are unintelligible. The problem is that we cannot argue for such counterfactuals in uncontroversial cases. For instance, we cannot argue that had particles arranged computer-wise failed to compose a computer, we would have believed this.“
Could you say more about what you have in mind here? I`m always a bit confused about people`s worries concerning counterpossibles since the fact that certain seemingly sensible and contentful ones come out trivial on the going semantics always strikes me as a spandrel of a not perfectly adequate model of their meaning.
Terence must feel like he’s battling a hydra.
So just in general: do readers think there is a good point to what Terence calls the mimicry project? I don’t want to contest the Blackburn exegesis, but I guess that project seems to me like an odd waste of time. Worse than a waste of time, actually, since I think what (some) realists say metaethicists should be trying to avoid saying.
I think something resembling the accommodation project is worthwhile. It’s a tricky question exactly what the to-be-accommodated includes. But I think it’s fairly central to contemporary expressivism that it not convict ordinary moral thought of massive error (as Mackie did).
Hi, Jack.
I agree with you. I regard the notion of “metaphysical” possibility as of little philosophical interest.
My concern was not that the counterfactual, “had the explanatorily basic moral truths been different, our beliefs would have been correspondingly different”, is vacuously true. (I take it to be false.) My concern was that analogous counterfactuals involving explanatorily basic truths – i.e., truths that link subvenient properties to supervenient properties — from other areas have equal claim to being false. In general, I think there are very few kinds of “metaphysical” necessities, F, which are such that, had the F-truths been different, our F-beliefs would have been correspondingly different. One could take this to show that we cannot even “account” for the truth of our ordinary object beliefs. I do not regard that conclusion as absurd. But certainly typical advocates of the “coincidence” argument against moral realism seem to.
(I should have added that “accounting” for the truth of our moral beliefs seems also not to be a matter of arguing that the probability that they are true is high. For either the probability is epistemic or it is objective. Whether the epistemic probability of our moral beliefs is high is what is at issue. So, suppose that the probability is objective. Then for any (explanatorily basic) moral truth, p, presumably Pr(p) = 1, given that such truths are necessary. Moreover, again, as genealogical considerations themselves illustrate, it may be that Pr(we believe that p) ≈ 1, because the probability of our having the (explanatorily basic) moral beliefs that we do is high. But, then, Pr(p & we believe that p) ≈ 1, by the probability calculus. Since, Pr(our belief that p is true) is just Pr(p & we believe that p), it may be that Pr(our belief that p is true) ≈ 1.)
Good god, I go skiing in the afternoon and this is what awaits me!
Let me try to address some of what folks have said, starting with Jussi and Justin. (Thanks for weighing in, both of you.)
Jussi’s second concern: I suppose expressivists could understand (a)-(c) to be normative claims. I can’t recall any passages in which they explicitly make this move, and I’d like to see a case made for it. But I suppose the move is available in principle. I’m not sure what making the move would get expressivism, however. Suppose, for argument’s sake, these claims were normative. Then what needs to be explained are these normative truths. If their being normative somehow implies that they require no explanation, then I think we’d need to hear why. (If it turns out that “normative truth” in the mouths of expressivists means something radically different from what it means in the mouths of realists, then I continue to harbor worries about in what sense expressivism would thoroughly mimic realism. I realize Jamie raises a worry about what extent expressivism is invested in mimicry in his second comment. I hope to get to that later)
Jussi’s first concern/Justin’s concern: As for tracking. I confess to have not reflected on this topic with the care it deserves. I think Justin is probably right that appeals to safety and sensitivity principles aren’t that helpful. (I myself find appeals to counterpossibles really puzzling in this context.) At any rate, that would leave at least the following options:
a. there is no satisfactory account of what it is for judgments to reliably represent/track necessary facts/truths. That is excellent evidence that there is nothing to be explained. The remarkable coincidence worries about realism and Thick expressivism come to nothing.
b. there is no satisfactory account of what it is for judgments to reliably represent/track necessary facts/truths but that’s simply because the phenomenon doesn’t admit of an illuminating gloss. Still, there’s a phenomenon to be explained.
c. we have no satisfactory account of what it is for judgments to reliably represent/track necessary facts/truths but we can be reasonably confident that we can identify one, given enough ingenuity. (Maybe we have good reason to hold that, given the nature of the facts, the relation would have to be in the neighborhood of a grounding/constitution relation.) In the meanwhile, a less than satisfactory grasp of what’s at stake will have to do.
d. What get presented as concerns about tracking are, at bottom, variations of Benacerraf-style worries: viz., how could we get necessary facts/truths in mind such that we can, in a fairly wide range of cases, form de re/predicative beliefs about them that exhibit various types of epistemic merits? What could the link between our mental states and these truths even be?
Some of these options look better to me than others. Those sympathetic with realism aren’t going to mind (a), since it gets their view off the hook too. I think my sympathies probably lie in the neighborhood of (d). That said, I think I also need to hear more about why learning that moral facts don’t ground/determine/constitute our moral beliefs wouldn’t cast doubt on the epistemic credentials, as Justin suggests, especially if that’s the only game in town for realists.
Thanks for your comment, Camil. Right, expressivists do like to focus on a particular range of data. I think it’s a good question whether, given the view’s aspirations and other commitments, it needs to widen its focus.
I think I disagree with the claim that expressivists wouldn’t deny that moral facts play a role in explaining how our moral beliefs could be accurate/mistaken. (I’m thinking of the Blackburn article in _Inquiry_ from 1999, in which he seems to say this and various places in which he says that expressivism must reject “thick” notions of representation; the 2001 _Philosophical Books_ symposium comes to mind). One reason for my disagreement is that expressivists go deflationary about facts, truth, representation, belief, etc., and deflationary views of truth, for example, tend to reject the claim that facts explain why propositions are true.
That said, if what you say is right, I’d like to know why moral facts enter into explanations of why some moral judgments/propositions are true but not into the psychological explanations of our moral judgments. I’d think the view would want to allow that we form some moral beliefs because we see that they’re true.
Hi, Terence. Thanks for the response. (I hope you had a great time skiing!)
I think that it would be bizarre if your epistemological argument *against* moral realism presupposed what is arguably an even more controversial kind of realism – e.g., realism about grounding. This would mean, for example, that empiricists like Mackie, Harman, or Kitcher would be unable to endorse the argument, despite advocating similar arguments themselves.
But set that aside. Could learning that the fact that p fails to, e.g., ground our belief that p “cast doubt on the epistemic credentials” of that belief (realistically construed)? Not if the quoted phrase means undermine, and the following principle, which I call Modal Security, is true.
Modal Security: Information, E, cannot undermine our beliefs of a kind, D, without giving us some reason to believe that our D-beliefs are not both sensitive and safe. (This is from my Justification and Explanation in Mathematics and Morality)
Learning that the fact that p fails to ground our belief that p gives us no reason to believe that that belief is not both sensitive and safe. Hence, it does not undermine that belief, if Modal Security is true. Of course, Modal Security could be false. But if it is, then there is a mystery: how could E obligate us to give up beliefs without threatening our verdict that they were “bound” to be true?
See also David Faraci’s post here: http://peasoup.typepad.com/peasoup/2015/02/normative-necessity-and-normative-knowledge.html
Thanks for your response, Terence.
Re deflationary facts and explanations of truth: I would think that a minimalist approach to truth and factuality allows one to adopt (or at least mimic) the realist’s truthmaking explanations. For instance, *that my belief about X is correct* can be treated itself in a deflationary spirit, as a first-order explanandum, in which case there is no reason for expressivists to avoid invoking (minimal) facts in accounting for it. There may well be passages in Blackburn that suggest otherwise, but I’m not sure how they fit with his overall view.
Re psychological explanations of beliefs: I agree that expressivists can allow for certain explanations of moral beliefs that appeal to the very facts those beliefs are about, e.g. I believe that burning cats is wrong because it is wrong. Gibbard in particular would be happy to accept such explanations as such, given his view on property identity. Blackburn may be more hesitant—see his *Just Causes*, where he suggests that in all such cases the supervenience bases of normative properties are doing the causal-explanatory work, and not the normative properties themselves. However, I think that expressivists must deny that a robust explanation of why we have normative concepts and the belief-forming tendencies that we do involves normative facts as what our concepts track or respond to. (The supervenience bases of normative properties may play a causal role in a narrow range of evolutionary pathways, but not in all scenarios in which people use normative concepts, e.g. Moral Twin Earth.) And here I’m not sure why a non-naturalist would disagree. Would non-naturalists want to claim that the best anthropological story about the emergence of normative thought involves a causal role for normative facts? I don’t see how this would fit with their standard thesis that normative facts are not causally efficacious.
Justin,
Thanks, that`s helpful. I take the point about probability to be that probability is—surprise, surprise!—a lousy account of accounting. I mean, after all, the probability of the truth of our belief in any recherché necessity will be diminishingly small on the probabilistic model given your translation.
My suspicion, for what it`s worth, is something between Terence`s b and c options, in addition to the foot-stomping claim that Modal Security is simply not true. I could try (try, mind) to bolster this by insisting that the sort of explanation of why our moral, mathematical, logical, etc. beliefs are safe and sensitive seems completely unsatisfying, presumably because the explanations do not connect up with the truth of the beliefs themselves. But cashing out this “connect up“ bit is exactly the problem, so I`m not sure what else there is here to be said. It is certainly true that the going accounts of this sort of “accounting“ are not satisfying. But, still, since it seems so clear that there is something to the disconnected-intuition, we must be on to something.
Anyways, thanks for the clarification!
Let me apologize for not addressing all the comments seriatim and say something briefly regarding Justin’s comments. I hope to get to the other comments as I can. (Yeah, excellent XC skiing to be had yesterday!)
First, I wonder whether focusing on questions about how to spell out the notion of tracking obscures from view what strikes me as a very plausible principle that’s driving the worries that Sharon and others raise. The principle is that for an agent to know some fact, that agent has to bear some appropriate relation to that fact. The worry that remarkable coincidence objections are raising, I think, are best viewed as being animated by this principle. If moral facts were as Thick expressivists (or quietists of some kinds) say, then it’s really difficult to see how they could satisfy this principle; the problem isn’t so much with their not being able to specify or explain what the appropriateness is but that their views appear to imply that there could be no relation (at least that’s not of a “Cambridge” sort). I acknowledge that working out the details of what the “appropriateness” is might be tricky. As I read Jack’s latest comment, he seems to be expressing something similar.
Second, I have reservations about Justin’s modal security principle. But my thoughts on this topic aren’t really my own; John Bengson (U. Wisconsin, Madison) has some work on this that I’ve found helpful and put pressure on the principle. So, I’d simply point those interested in the topic in John’s direction.
Third, I wonder, Justin, about the first point you raise. If grounding/determination/constitution were the only game in town, I think that would no more debar Mackie from pressing the objection any more than it would debar him from developing a version of the problem of evil given his commitment to the error theory.
Hi Jamie,
I think you raise a really important point when you ask about expressivism’s commitment to the simulation project. As you point out, we need to distinguish what expressivists actually say (under a charitable interpretation) when developing their view and what they should say.
As for what expressivists do say, I think there’s really strong textual evidence that that they commit themselves to the simulation project in which the mimicry or emulation is thorough and extensive. This project seems to me to lie at the heart of quasi-realism. It’s what makes the creeping minimalism worry seem like a real challenge. I take it that the reason expressivists commit themselves to the project is that, for all their reservations about realism, they think realists get some important stuff right: there are moral facts, there are moral truths, these facts and truths are objective in the sense of being attitude-independent, we have moral knowledge, and so forth. The trick is to affirm all these things without (1) engaging in “devious reinterpretations” of moral thought and discourse (G. Rosen’s nice phrase) and (2) without committing expressivism to robust moral beliefs and facts.
As for what expressivists should try to mimic, I think that’s something about which we need to hear more from the expressivists (and their sympathizers). As you point out in your _Analytic Philosophy_ article, Gibbard begins to say some things about this, distinguishing vast realism from tempered realism. But, as you also point out, the discussion isn’t that helpful, as it’s difficult to see what the distinction comes to.
Hi Nick,
Thanks for your observation. I agree: what you quote doesn’t sound like a metalinguistic/semantic claim. But I think there are lots of passages in which expressivists do go this route. For example, Jamie quotes a passage in his _Analytic Philosophy_ paper on the topic in which Gibbard goes metalinguistic. And it’s the route Blackburn takes when addressing the objection that expressivism commits itself to the claim that moral facts are objectionably attitude-dependent. Moreover, when Blackburn says what he does in the passage you quote, I think it’s always a live question how to understand what he says. Finally, I’d say that when I pair the passage you quote with others in which the claims are clearly metalinguistic/semantic, it doesn’t ease my worry that expressivists sometimes slide between two different views (Thin and Thick)!
Both you and Jussi propose that expressivists could think of reliability as a normative notion. I’m just not sure what that gets them or why it matters that normative facts can only be assessed from within a normative framework.
It’s a good question why a third-factor story isn’t available. I didn’t offer any argument for the claim but my worry is that to make the story work, expressivists would need to identify some entity that can (a) explain the moral facts and (b) explain our moral judgments but ensures that (c) the moral facts don’t explain our moral judgments. At various places, Blackburn invokes norms of correctness and Mike Ridge invokes “acceptable standards.” They might be good candidates for being “intermediary” facts.
Suppose they are. This approach would then posit a three-term relation, with moral judgments, intermediary facts, and moral facts being the values of the three terms of the relation. Suppose, for argument’s sake, we can easily grasp the intermediary facts. We still need to explain why for each moral fact M, there is an intermediary fact F that is invariably correlated with M. Holding that there is no explanation of this correlation is not an option, since that amounts to simply positing another type of unexplained coincidence in order to explain what we initially thought was an unexplained coincidence. Moreover, holding that the moral facts explain the correlation is also not an option, as the moral facts could no more explain this correlation any more they could explain the invariable correlation between our moral beliefs and the moral facts when our moral beliefs track the moral facts. For if moral facts were to explain either correlation, then they would be robust. The remaining option would be to explain the existence of the moral facts by appeal to the intermediary facts themselves. However, I wonder if this is a genuine option either. For, presumably, if there were intermediary facts, then they would be normative facts and, hence, no more robust than the moral facts. But if they are not robust, then they could neither explain the fact that we track the moral facts nor the existence of the moral facts.
Anyway, that’s my worry.
Thanks, Jack.
I think that the concern you raise about recherché necessities can be handled by conditionalizing on the probability that we hold some belief on the matter. But this would do nothing to vindicate the view (of Street et al) that the probability that our moral beliefs are true is low, given moral realism.
Thanks, Terence. I would like to distinguish two questions.
(1) Can we “explain the reliability” of our beliefs of a kind, D, in a pretheoretical sense of this phrase?
(2) Can we “explain the reliability” of our beliefs of a kind, D, in every sense which is such that the apparent impossibility of explaining their reliability undermines them (realistically construed)?
I suspect that the answer to (1) is “no”, whenever D is an apparently a priori domain. We seem to seek a connection between, e.g., moral states-of-affairs and our moral beliefs that mimics causation, but need not be causal. Neither explanatory indispensability nor grounding (and like ideology) seems to satisfy this constraint, because, unlike causation, neither is predictive of sensitivity or safety.
However, an answer “no” to (1) strikes me as of limited interest if the answer to (2) is “yes”. In that case, either we *can* show that the truth of our moral beliefs (realistically construed) is “no coincidence”, or we cannot, but knowledge of this does not undermine them (so construed). Either way, coincidence arguments would fail, and realists (or “thick expressivists”) would be given no reason to change their views. (The idea that realists would be given reason to lower their credences, but not below the threshold for belief, strikes me as unexciting.)
What if one could argue that answering “no” to (1) required answering “no” to (2)? That would be interesting! But, if what I have said is correct, then arguing for this conclusion would require arguing against Modal Security. To the extent that it is hard to see how Modal Security could fail, such an argument would illuminate the nature of undermining. (I will talk to John Bengson about this.)
(I wasn’t sure what you meant by the antecedent, “If grounding/determination/constitution were the only game in town…”. Could you clarify?)
Terence,
realists get some important stuff right: there are moral facts, there are moral truths, these facts and truths are objective in the sense of being attitude-independent, we have moral knowledge, and so forth.
On textual evidence: I suppose you mean Blackburn, but he’s changed his mind a few times. If you mean Gibbard, I’d be quite interested to see what you have in mind.
I don’t think your mimicry project lies at the heart of quasi-realism. The (proper) project is to make good sense of what people say in moral discussion and think in moral thought. It should be no part of the project to mimic what metaethicists of any stripe say. Most metaethicists, after all, are wrong.
In ordinary moral conversation people talk about moral facts, say of some moral claims that they are true, probably do not say that anything whatsoever is mind-independent but do assert and believe some of what I call Independence Counterfactuals (“x would be wrong even if… were not …”). It’s important to be able to make sense of these, lest expressivism turn out to be an error theory.
Minimalism is very helpful in this project, but then turns out to Creep, generating a new problem.
Does this summary give you pretty much what you wanted? Or did you want Gibbard to be able to assert every sentence of The Normative Web? (Except the acknowledgments.)
Hm, my blockquote tag failed. Well, the paragraph that doesn’t look like a paragraph is a quote from TC’s earlier comment.
Jamie,
Actually, I have both Gibbard and Blackburn in mind. They’ve both changed their minds on various issues, I take it. Blackburn, however, usually leaves it as an exercise for the reader to figure out what the changes are.
I think we disagree deeply about what the quasi-realist project is. Apologies in advance if the discussion now takes a Talmudic turn.
When I open up THTL, I read statements like these:
“Much of what non-naturalists say is right…. Simon Blackburn calls the program we share _quasi-realism_: from a basis that excludes normative facts and treats humanity as part of the natural world, I explain why we would have normative concepts that act much as normative realists proclaim…The force of all this, if it works, is to confirm the hypothesis I’m exploring. Suppose that (i) normative realists are right about how normative concepts act, and (ii) my hypotheses has a consequence that normative concepts act this way” (xii).
As I read it, this does not say the quasi-realist project is simply to make sense of what people are doing when engaging in ordinary normative thought and discourse. Such thought and discourse, after all, could be radically different from what realists say about such thought and practice. Rather, as I read it, the passage says that moral concepts work pretty much as (nonnaturalists) realists say but that expressivists can explain their workings without committing ourselves to everything that realists commit themselves to when they explain their workings.
Here’s another passage from Gibbard:
“With vast normative realism contrasted with a halfway, tempered normative realism, we can ask which kind a quasi-realist should aspire to mimic. The answer is clear enough. Once we convince ourselves that we are the products of the evolution and ecological dynamics of our species with its special cultural histories, we can’t take a vast normative realism seriously. The only credible candidate for emulation is a tempered normative realism…What quasi-realism mimics is not tempered realism as a whole, but tempered realism in all but one respect…We can’t, though, mimic the claim that understanding normative properties and relations as objective matters of fact is basic to explaining how judgments of wrongness work” (2011, 45-46).
Putting aside the difficulties in parsing the difference between vast and tempered realism, Gibbard strikes me as explicitly stating that quasi-realism aims to emulate a metaethical position, viz., tempered realism, and that view only up to a certain point.
I realize that one could dismiss passages such as these as stray or claim that quasi-realists regularly misrepresent their own position. I don’t read them that way. Under my reading, quasi-realism is invested in the simulation project (it’s quasi-_realism_, after all). Whether they should be so invested is a different issue. But when engaging with the view, I think the place to start is how they actually characterize its aims and, as I say, I think there’s pretty strong evidence for the claim that the view endeavors to mimic what nonnaturalist views say about what I’ve called the core moral data.
As for Gibbard and _The Normative Web_: I’d settle for a dramatic reading of the text, including acknowledgments.
I am a quasi-realist. I tend to characterize quasi-realism’s overriding aim as making sense of the realist-sounding things that ordinary people say and think. Crucially, something can *sound* realist – that is, like the sort of thing only a realist could make sense of – without actually being realist – that is, the sort of thing only a realist can in fact make sense of. The target, then is ordinary thought and discourse – but it is conceded at the outset that much of this discourse *sounds* realist in spirit. That is quite different from starting out with some worked out theory (which one?) and trying to emulate all or most of what that theory says. The project I have in mind is much closer to the one Jamie lays out above. We can debate the extent to which those and other passages in Gibbard depart from that aim – for example, the ‘much’ in ‘much of what non-naturalists say is right’ can be taken to be doing more work than your reading allows.
But in any event, there are two projects here – the one you lay out and argue can be found in Gibbard and Blackburn, and the one Jamie suggested and the one that I take myself to be pursuing. I think there is a good case to be made that this project, rather than the one you are targeting, is a more philosophically interesting and promising project. Do you disagree, and if so, why?
Incidentally pretty much precisely this issue about which of these two ways to construe quasi-realism frames one of the questions Mark Schroeder gives as an exercise to his readers in his textbook on non-cognitivism (see p. 168, question 10).
Hm, I think AG’s words can be construed either way.
But anyway (following more along the lines of what Mike says), let’s imagine for a moment that we have a really precise and accurate version of how normative concepts work according to non-naturalist realists, and we also get a precise and accurate version of how normative concepts work in the actual world of human beings who possess and use these concepts. And, we find, these are not the same. (Which is looking more likely; see Sarkissian, Park, Tien, Wright, and Knobe, 2011)
So now there are two quasi-realist mimicry projects. But one of them seems like a bizarre and prima facie pointless exercise, while the other has on its face a serious and important motivation. I hereby express my plan to pursue the second if in Allan Gibbard’s shoes.
In “Knowing What to Do, Seeing What to Do,” (2002) Gibbard seems to commit to mimicing something in the area at issue. He discusses, “the problematic final clause in a definition of knowing, the “no fluke” clause,” as part of his account of how we know normative truths by intuition. Earlier in the article he argues quasi-realists should want to give such an account.
Any way, he connects the no fluke clause to reliability and then argues that reliability is frought with ought/planning judgments. He concludes this sketchy discussion, though, by admitting that there is more to do to fill out the no fluke clause and he suggests we could do that by mimicing accounts of knowledge of “how things stand.”
Right after mentioning the no fluke clause, Gibbard mentions that knowledge claims license reliance on the process, not just the verdict, so he seems to understand the the call to fill out the no fluke clause *in a theory of knowledge* as a call to underwrite that sort of license to rely.
Terence, I am having trouble seeing why Gibbard would need to answer the explanatory question you are pressing in order to underwrite the relevant licenses or develop his theory of intuitively based normative knowledge. On the other hand, I agree with you that a theory is more intellectually satisfying if it can back or elucidate the thought, for example, that I know that recreational slaughter is wrong because recreational slaughter is wrong. But is this just because I lean internalist in epistemology? Or do you think there is some neutral demand regarding knowledge that should push Gibbard to think that a theory of normative knowledge is better if it accommodates or elucidates the thoughts at issue (e.g. that I know that recreational slaughter is wrong because recreational slaughter is wrong)
I am way out of my areas here so I hope I am not just confused – just following the call for half-baked thoughts.
Justin,
Agreed, again, though if you conditionalize on the one, you should conditionalize on the other, given that you’re trying to “translate” “our belief that p is true”. Anyways, clearly the conditional patch will not favor Street’s claim and it’ll avoid my cranky pedantic point above. I tend to think this just shows, again, the inadequacy of using such a model in this case. But I should stop agreeing with you (i.e. stop threadjacking).
Just to agree a bit more with folks, I agree with Jamie and Mike about the sensible versions of quasi-realism. I discuss this a bit in my imprint paper when I talk about the distinction between hermeneutic and revolutionary expressivist views (not the same as quasi-realist views, of course, but close enough for this point.) Mike and Sebastian have a nice paper on it as well! The interesting question for the quasi-realist is what is involved in making sense of ordinary objective-seeming discourse. That is, at what point we draw a line between rational reconstruction and revolution.
Hi Mike,
I thought you might chime in at some point! Thanks for the pointer to the Schroeder discussion.
Here are a few thoughts in response.
First, my aim in airing the challenge to expressivism was not to claim that all versions of (quasi-realist) expressivism are subject to it. I claimed only that some prominent versions of the view are. I think that, under a plausible reading, Blackburn’s and Gibbard’s views are among these prominent versions. But, I’m happy to concede that, in principle, some versions of expressivism may side-step the worry at issue entirely.
Second, I worry that the terminology that expressivists tend to use is deeply misleading, at least if the best version of the view eschews mimicking realism. As I see things, we should distinguish: (a) the core moral data (b) a given metaethical theory (realism, expressivism, etc.) that endeavors to accommodate/explain the core data and (c) how the metaethical theory endeavors to accommodate/explain the data. The data, as I think of them, are largely pretheoretical. So, it’s difficult for me to make sense of the claim that the data are “realist seeming” or anything of that sort. Admittedly, realism is typically fashioned in such a way that it aims to accommodate/explain these data by appeal to its core theses. But, so far that I can see, the relationship between realism and the data is not more intimate than this. In fact, expressivism and realism might agree almost entirely on what the data are that need to be accommodated/explained.
So, it may be that the best versions of expressivism are committed to only this: accommodate/explain the core data in such a way that avoids a commitment to the error theory, various forms of relativism/subjectivism, and realism. If the view’s commitments very significantly diverge from realism at certain points, then so be it.
If this is the way expressivists want to go, then I think there’s a lot of explaining to do that, to my knowledge, hasn’t yet been done. We need to be told what, in their view, the core data are. Suppose the core data include moral requirements, for example. Then it’s difficult to see how the view accommodates/explains the data by appeal to metalinguistic/semantic theses about what it is to say that there are such requirements (or what it is to say that it’s true that there are such requirements). Suppose, in contrast, the core data include only the appearance that there are such requirements. Then we need to know why appeal to metalinguistic/semantic claims accommodates/explains those appearances. (This makes me wonder whether the explicitly projectivist dimension of the project in Blackburn 1984 might be much more important to expressivism than the current discussion makes it seem.)
In short, I wonder that even if expressivism rejects the simulation project, related worries arise.
Hi Terence,
I am more optimistic than you about dividing putative data points as between those that seem realism-friendly and those that don’t. When I say ‘realist-seeming’ I don’t mean realist-seeming *to the folk*, by the way, which may disarm your point about how the data points are largely pre-theoretical. Granted the data themselves are intuitive, but we can then ask “the experts” which of those data points seem on the face of it to be more easily explained within a realist framework and those which don’t seem to fit so well into such a framework (and those which are neutral). The folk may also not have clear intuitions about which first-order intuitions are consequentialist-sounding and which aren’t but that doesn’t stop the experts from having well grounded opinions about this.
To illustrate what I mean: I suspect if you gave philosophers the following list of putative data points that you would get a *lot* of agreement as to which ones are “realist-sounding” and which aren’t:
That moral judgments can be true or false
That some of them are true
That their truth can outrun or diverge from our judgment (even our best judgment)
They can figure in rational inferences
We can be more or less certain of them
That they can motivate us independently of any independently existing desire.
That people disagree quite deeply about them, where this outruns any mere factual disagreement.
[one way to help get a grip on how we divide these is to construe the distinction between realist-sounding and non-realist-sounding data points at least in part in terms of features which make the judgments sound like ordinary beliefs with moral contents and those which make it sound more plausible that they are desire-like states]
That said, there is a no doubt a way of understanding ‘core data’ on which quasi-realists certainly can and should accept a commitment to explain the core data. I don’t think that there actually are moral obligations is one of those core data, though, and for a principled reason. Given quasi-realism it follows immediately that whether there are any moral duties, and if so what they are, is trivially a first-order question – one that you can answer only by adopting some attitudes. So it would flagrantly beg the question against quasi-realism to assume it had to *explain* why we actually have moral duties. I see that you actually say ‘accommodate/explain’ but these are quite different. Insofar as ‘accommodate’ just means something like ‘explain the intelligibility of’ then quasi-realists do aspire to do this, and in my view do a good job. So I don’t see any big new problems arising from the one example you give here.
And in any event, this would entirely side-step any problems arising from the “mimic the realist’s theory” research program you originally had in mind. I take it that you agree, or anyway don’t want to argue against the thesis, that this research program is less philosophically interesting than the one Jamie and I have in mind (and the one I still think is what both Gibbard and Blackburn really have in mind, actually but I conceded that some passages are open to both readings).
PS Terence is this what you’ll be discussing at the Southampton workshop in June? Just curious now since it is very much on topic!
Hi Terence,
Thanks for your post, as it gives me a chance to point out something about the relationship between expressivism and (quasi-)realism that seems to have gone almost entirely unappreciated (we talked about this a bit at Madison, if you recall).
Thin expressivism is just explanatory, aiming to explain states of mind, attitudes, etc. There is no practically normative commitment expressed by thin expressivist claims (at a minimum, such explanations are just as consistent with abolitionism about moral discourse as are error-theoretic claims). By contrast, what you are calling thick expressivistic claims do express normative commitments, namely practical commitments to the sorts of states of mind that the thin expressivist is in the business of explaining.
So, in contrast to some of the commenters, I think there is a fundamental difference between thin and thick expressivism. The first is an explanatory project and the second (normally called quasi-realism) is a normative project or commitment. Especially in the case of moral expressivism, these can come entirely apart. One can be a thin moral expressivist while wholeheartedly rejecting moral discourse (insofar as it is committed to attitude-independent reasons). The fact that many expressivists have been trying so hard to ‘earn the right’ to talk like moral realists has obscured the fact that this is a fundamentally normative project. One could just as well employ thin expressivistic claims to justify *not* talking like a moral realist. In either case, one will employ the thin, explanatory claims to get traction on the fundamentally normative question of how to think and talk. Blackburn and Gibbard rightly point out (along with other non-expressivists like Richard Joyce) that the attitude-independent character of moral discourse serves a committing function. This is offered as a kind of justification for employing moral discourse. One could roughly agree with their explanatory claims (and their deflationary gambit) while drawing very different normative conclusions (to see how this would go, please see my “Breakdown of Moral Judgment” in Ethics).
In short, thin expressivist claims can be made by a non-participant in moral discourse, just as one could explain the states of mind expressed by religious claims without being a participant in religious discourse. By contrast, thick normative/moral expressivist claims must be made as a *participant* in normative/moral discourse. In my view, the failure to recognize that these are fundamentally different stances, and that fundamentally different norms of justification properly apply to thin vs. thick expressivistic claims is a source of deep and abiding confusion when it comes to evaluating expressivism and quasi-realism in particular (the latter being a practically normative project, totally divorceable from the expressivists’ explanatory project).
The expressivists’ explanatory accommodation project has often been run together with the normative project of justifying a commitment to employing moral discourse, including and especially its commitment to attitude-independent reasons. These two different kinds of accommodation project are rarely distinguished in such a way that allows one to see how easily they can come apart.
In your paper “Expressivism’s Deflationary Gambit”, I think you run these two projects together. You write that “The Expressivist Explanatory Strategy” is “to begin with [an] understanding of the way in which moral concepts work and, on the basis of this understanding, to account for … the existence of moral properties, propositions and facts”.
But this very common understanding of what the expressivist explanatory strategy amounts to is, in my view, deeply mistaken. The *explanatory* project does not have to account for moral truths any more than a strategy to explain religious discourse must account for religious truths. For reasons I will refrain from explaining, the deflationary gambit (if successful) guarantees that the question whether to accept that there are moral facts, properties, and so on boils down to the fundamentally normative question whether to be a participant in moral discourse. Not only can we answer that question with an enthusiastic ‘No’ while being a committed (thin) expressivist, but we can employ the expressivists’ explanatory resources to help explain why we should *not* employ moral discourse (again, see “Breakdown” for a better sense of how this would go).
Many have wondered whether and how expressivism is different from non-naturalist realism. But expressivism as an explanatory project has nothing whatever to do with realism or antirealism. Quasi-realism, on the other hand, employs expressivistic (explanatory) resources in the service of justifying ordinary ‘realist’ discourse.
More generally, I think one of the main attractions of an expressivistic approach (and the deflationary gambit) is to help us see that whether to accept claims like (A) – (C) and (a) – (c) are fundamentally normative questions about how to construct our normative discourse. There is no obvious conflict between expressivistic explanations of our existing discourse and any of a variety of answers as to how we ought to conduct or construct that discourse in the light of these explanations–including whether we should accept claims like (A) – (C) and (a) – (c).
Thank you again for your post and please forgive me if I sound overly strident or critical. The tight association of expressivism and (quasi-)realism is a source of abiding frustration for me.
Hi Mike,
I was thinking of doing something on the error theory instead.
For reasons not unrelated to those raised by Eric’s post, I have sometimes wondered if the project which goes under the heading ‘quasi-realism’ should not be split into two projects, one of which is more fundamental and which would be more appropriately called ‘quasi-descriptivism’ or some such, since ‘realism’ implies a success theory, and the other of which would be a project in first-order normative theory to be carried out against the backdrop of the meta-theory on offer. Which is to say I’m broadly sympathetic to the distinction and the claim that it is often unhelpfully elided. I guess the upshot here is you could well be a quasi-descriptivist in this sense and also be a radical sceptic about moral knowledge – and that combination of views would clearly not be touched by this line of argument.
This has been a super interesting discussion for me. Thanks to all who have jumped in.
Here are additional thoughts in the neighborhood of Mike’s and Jamie’s comments.
A. On reading the canonical quasi-realist texts: in short, I think it’s difficult. I’ve tried to identify a charitable but faithful reading of the texts. I realize that Jamie and Mike think it’s not the best reading in the sense that it would commit expressivists to things that, on reflection, they might not want to commit themselves to. That said, I continue to think it is a faithful reading. The language of mimicry, after all, is all over the canonical expressivists texts. I think it’s very difficult to make sense of this language if expressivists are interested in only the accommodation project plus avoiding error theory/relativism/realism. As best I can tell, it wouldn’t make any sense to talk about mimicking the core data. I’ll add that I think the reading I’ve advocated, which has expressivists invested in the simulation project, would not commit expressivism to “a bizarre and prima facie pointless exercise.” If one thought that realism is correct about a lot – as Gibbard explicitly says on a straightforward reading – then you can see why engaging in the simulation project makes sense. Perhaps think of things this way: realism is explicitly designed to accommodate/explain the core data. To the extent that they succeed in this – and I agree that there can be reasonable disagreement about this – that would make mimicking much of what they say worthwhile.
B. On core data: Mike, my thought would be that the core data would be extracted from ordinary moral thought and practice and largely pretheoretical. I don’t think that if the extraction process were to yield core data that states that moral requirements exists, this would beg any questions against the expressivists (or the error theorists, for that matter). If, for example, reflection on the practice of playing chess were to yield as part of the core data the existence of the rules of chess, I don’t think we’d beg any questions against the rule-skeptic. The rule skeptic might have various things to say in reply, of course, that might justify her approach.
C. On realist-seeming data: I continue to find it strange to speak of realist-seeming data. I agree that some of the data appears particularly amenable to a realist accommodation/explanation, but that seems to me something genuinely different from the data being realist-seeming. (I also agree that accommodation and explanation are distinct; that’s why I included both.) Moreover, if we can make sense of the idea, all (or nearly all) the data that seems realist-seeming seem equally “constructivist-seeming,” “relativist-seeming,” “error-theorist seeming.” Maybe that’s fine. But I do think that characterizing the data as seeming-in-a-metaethical-way is not the way to go.
D. On first order questions: Mike, you write:
“Given quasi-realism it follows immediately that whether there are any moral duties, and if so what they are, is trivially a first-order question – one that you can answer only by adopting some attitudes. So it would flagrantly beg the question against quasi-realism to assume it had to *explain* why we actually have moral duties. I see that you actually say ‘accommodate/explain’ but these are quite different. Insofar as ‘accommodate’ just means something like ‘explain the intelligibility of’ then quasi-realists do aspire to do this, and in my view do a good job. So I don’t see any big new problems arising from the one example you give here.”
I confess that I find it hard to parse what this passage says, although it’s the sort of thing I hear expressivists say frequently.
a. it could mean that when realists and other metaethicists speak of the existence of moral duties, facts, truths, etc., then they’re simply engaging in first-order normative thinking, which consists in the expression of attitudinal states.
If this is the claim, then I think realists (and constructivists, error theorists, relativists, etc.) must and should reject it. I adverted to this rejection above when I assumed that realists can say what they mean.
b. It could mean that when expressivists (and only expressivists) speak of the existence of moral duties, facts, truths, etc., then they’re simply engaging in first-order normative thinking, which consists in the expression of attitudinal states.
That may be so. The unhappy result is that now realists/constructivists/error theorists/relativists, on the one hand, and expressivists, on the other, are talking past one another in a very radical way. (Another concern is one that we talked about several years back and that is whether the approach generalizes to all the normative domains (chess, baseball, prudence, etc.) I won’t rehearse the concern again, but it may be worth thinking about.) If this is the right way to understand what expressivists are saying, at least we know what’s at issue between expressivism and realism. I’m not really sure how one would go about attempting to resolve it.
c. It could mean that many of the claims that express the core moral data have to be given an expressivist treatment. The claim that there are duties, for example, receives an expressivist treatment but the claim that moral judgments have the marks of belief do not.
If this is what’s going on, then realists/constructivists/error theorists/relativists, on the one hand, and expressivists, on the other, would deeply disagree about what the core data are and, hence, what must be explained. The disagreement would mean that we’d be talking past one another in the sense that what realists are trying to accommodate/explain is fundamentally different from what expressivists are trying to accommodate/explain.
E. On whether problems remain: I think that depends on how we resolve the issues in D. If the core data include either duties or the appearances thereof, then I think there are some genuine issues that need to be tackled in a way that doesn’t consist in simply telling us what it is to say/think that something is a duty, a moral fact, etc.
Eric,
Hear hear. I find the same thing frustrating. There is a non-trivial gap between expressivism and quasi-realism, which is massively under appreciated due to the fact that many expressivist are also quasi-realists. On the other hand, I don’t understand this.
“More generally, I think one of the main attractions of an expressivistic approach (and the deflationary gambit) is to help us see that whether to accept claims like (A) – (C) and (a) – (c) are fundamentally normative questions about how to construct our normative discourse. ”
While I may agree that thick expressivists take on normative commitments in accepting these claims, I hardly see that whether or not to accept them are fundamentally normative questions about how to structure our discourse. Rather, they seem questions which carry normative baggage. Anyways, clarifications would be appreciated.
Terence,
Just to remind you: I said it would be a bizarre and prima facie pointless exercise on the assumption that the way normative concepts work according to non-naturalist metaethics is not the same as the way normative concepts work as actual people employ them.
Do you mean to disagree with that?
Eric,
I agree that the expressivist component of quasi-realist should be more clearly separated from the normative endorsement of realist-sounding commitments whenever we talk about these issues. But I don’t think Terence’s distinction between thin and thick expressivism maps onto the distinction between expressivism as such and quasi-realism. He says:
“Thick expressivism goes further [than thin expressivism], maintaining that to commit oneself to (A) is to commit oneself to there being moral facts (albeit not robust ones).”
So, as I understand it, thick expressivism adds to thin expressivism a semantic/metalinguistic thesis (that when we say “there are moral facts” we do thereby commit to moral facts, understood in a deflationary sense), not a normative commitment to the existence of moral facts.
Now, I do have some worries about what thin expressivism amounts to: does it fully reject the ideology of truth and factuality, even understood in a minimalist sense, like old-school versions of expressivism? And is Terence’s claim that, when quasi-realists have responded to epistemological challenges by appealing to their semantic account about attitudinal states, they have implicitly relied on such a stripped-down version of expressivism?
Oops, sorry, Jamie. I quoted from the wrong post. In an earlier post, you wrote that the simulation “project seems to me like an odd waste of time. Worse than a waste of time, actually, since I think what (some) realists say metaethicists should be trying to avoid saying.” It’s that claim I meant to address.
But, no, I don’t mean to disagree with what you say in the later post (on the assumption that expressivism isn’t trying to reform moral discourse in a certain way).
Hi again Terence,
On (A), I don’t think we are likely to settle this exegetical question about Blackburn and Gibbard here. There are lots of passages that could be cited by both sides, I think, and many are open to multiple interpretations. In any event, as a no doubt “non-canonical” quasi-realist, though, I think I have been pretty clear and consistent in construing the view in the way I did here.
On (B) I think what is potentially question-begging is assuming that a purely meta-ethical theory must explain some putatively first-order data point, where what counts as first-order or meta- depends on which meta-theory is correct. The chess analogy might be instructive here. A meta-theory of chess will presumably explain in virtue of what such-and-such rules count as the rules of chess and that sort of thing (appeals to relevant conventions will play a role here). Even though it is a fact that certain moves are terrible, it would be inappropriate to insist that a meta-theory of chess explain why they are terrible. That just isn’t within the intended scope of the theory. What the meta-theory should do is simply explain what we are up to when we make judgments about moves being good or bad, etc. That seems analogous to the issue of there being moral duties to me.
On (C), there is perhaps a terminological problem with ‘realist-seeming’. What I am really after is ‘cognitivist-seeming’ – which includes all of the other theories you mentioned since they are all species of cognitivism as standardly understood. With that terminological switch, I think the basic approach is sound. This is not unrelated to my suggestion above that the view should really be called ‘quasi-descriptivism’ or perhaps ‘quasi-cognitivism’ if you like. In fact, it is very closely related to that! To that extent, I have probably been guilty of being caught up in the standard terminology in the literature a bit.
On (D), I actually think a lot can be said in favour of your (a), but I don’t think I need that assumption here, which I realize is controversial. Rather, my point was a dialectical one. Expressivsim is a meta-theory and so only sets out to explain certain sorts of data. Insofar as a data point is an entirely first-order data point, expressivism in this sense simply doesn’t purport to explain it. Since it would beg the question to assume that there are moral duties *in some sense of ‘there are moral duties’ that is incompatible with expressivism* the expressivist is under no prima facie pressure to explain the existence of moral duties – as opposed to making intelligible our thought that there are such duties. The analogy with a meta-theory of chess versus a first-order theory of how to play chess well is perhaps useful here – see above.
On (E) – yes, this does depend on (D) – there, something we agree on!
Oh and Jamie beat me to it on the point he just made! The qualification is essential to his original thesis.
Jack,
Thanks for sharing my frustration :). Let me do what I can to clarify what I say in the quote you pull. The way I put it is misleading, or at least unclear.
On the one hand, we can ask questions about what sort of first-order practical or normative attitudes to have (e.g., related to recreational slaughter). On the other hand, we can ask questions about how to express, conceive, and justify these attitudes. These are distinct but related kinds of normative questions.
What I meant in saying “whether to accept claims like (A) – (C) and (a) – (c) are fundamentally normative questions about how to construct our normative discourse” is that whether to express our attitudes, plans, values and the like in terms of (moral) facts or as instances of (moral) knowledge, or in representational terms, etc. is a fundamentally practical, normative question.
Part of why expressivism can help us see that whether to accept that there are any moral facts is a fundamentally practical, normative question is that expressivists explain our talk of moral facts, duties, knowledge, etc. in ways that do not advert to such facts, duties and knowledge. Rather, talking and conceiving of the normative realm in these ways is explained in fundamentally practical terms.
For example, Blackburn and Gibbard both explain our (implicit and explicit) talk about attitude-independent normative facts in terms of the role such talk plays in sustaining practical commitments in the context of temptations to act contrary to one’s own commitments or plans. Blackburn employs this basic idea to defend moral quasi-realism, saying that it would be bad to think about the wrongness of torture as dependent on our attitudes toward it–it would make us less committed to opposing torture. That is a practically normative argument for accepting that there are moral facts, understood as normative facts that are independent of one’s own attitudes. (The reason why this sort of move seems ‘schizophrenic’ to many people is that it is a fundamentally subjectivist justification for employing objectivist language and concepts–like Joyce’s motivational fictionalism without the error theory).
On the other hand, if there are serious practical downsides to thinking in terms of attitude-independent normative facts, then those considerations can form part of a normative argument against accepting that there are moral/normative facts, knowledge, etc.
The deflationist/expressivist approach helps us see that the question whether we should accept that there are moral facts cannot be addressed, even in part, by asking a prior question whether there are *actually* moral facts (just as we cannot ask whether we should accept that recreational slaughter is wrong by asking a prior question about its wrongness). In the most general terms, whatever sorts of resources we can draw on to justify first-order normative claims, we can draw on to justify claims about how to generally conceive, express, and justify those first-order claims. In other words, it’s normative all the way down.
I hope that’s helpful, but experience has taught me to keep my hopes in check in this regard :).
Camil,
Thanks for your comment. Your question brings out how difficult (and important) it is to separate the different kinds of claims I referred to in my first post. I think Terence characterizes thick expressivism in a way that doesn’t separate them well enough (but please correct me if I’m wrong, Terence!). He first characterizes thick expressivism in the way you say, as merely the view that moral claims commit their makers to moral facts. But then he moves directly to:
“Thick expressivism endorses both:
(B) Moral judgments often accurately represent the moral facts; and
(C) When moral judgments accurately and reliably represent the moral facts – that is, “track” the moral facts – these judgments are often cases of moral knowledge.”
These claims go far beyond the initial characterization, which implies nothing about whether there are any moral facts, duties, knowledge, etc. We can be a thick expressivist in the first sense without accepting (B) or (C).
By analogy, I accept that when people make claims about what God wants, they thereby commit themselves to the existence of God (albeit not necessarily in a robust sense). It does not follow from this that I accept or deny that divine command claims ever accurately represent or track any facts or are cases of genuine knowledge. In my view, to accept that divine command claims can be cases of knowledge requires accepting the legitimacy of divine command discourse (In my view, whether to do so, and in what contexts and senses, is a practically normative question).
In general, I think we can elucidate what sorts of commitments people are making in divine command discourse, moral discourse, taboo discourse, or any of various kinds of normative discourse without making any commitments at all about whether any of these claims are accurate, cases of knowledge, or the like. Whether to make the second sort of commitment *in general* is, in my view, a fundamentally normative question about whether to be a participant in that discourse. Whether to make *specific* normative commitments within a given normative discourse is a first-order question within that discourse.
I think the reason that this transition (from an explication of commitments to an implicit acceptance of the legitimacy of those commitments) obviously doesn’t work in the case of divine command discourse, but is so hard to even see in the case of moral discourse is that pretty much everybody is (and thinks they should be) a participant in the latter discourse. We therefore tend not to notice (much less be jarred by!) the slippage from an explication of the commitments of participants within moral discourse, to an expression of the legitimacy of those sorts of commitments.
I think it’s worth reading the relevant passage again, substituting ‘prohibited by God’ for ‘(morally) wrong’ and then trying to feel the sense of jarringness when moving from the first characterization of thick expressivism (now a claim about what people are committed to when making divine command judgments) directly to the idea that thick expressivists about divine commands must endorse (B) and (C), understood as claims to the effect that divine command judgments often accurately represent facts about what God commands (albeit not necessarily in a robust sense).
I wish I had time to explain myself more fully, but I should really be using that time to write these sorts of things up for publication :).
Hi Eric,
Thanks for your post – lots to digest there. I agree that expressivism and quasi-realism are distinct projects. I hope I didn’t run them together in the paper you cite or in this post. I did qualify what I said both here and in that paper by noting that I was addressing expressivism in its most sophisticated forms, which is quasi-realist in cast. As best I can tell, the most sophisticated versions of expressivism have invariably been quasi-realist; it seems legit to focus on those even if it’s possible to develop the expressivist component without the quasi-realist component, which commits us to the legitimacy of thought/talk regarding moral facts, truth, knowledge, etc.
That said, I’ll have to read your paper to see whether failing to distinguish Thin from Thick expressivism (in your sense of “thin” and “thick,” which is different from mine) creates mischief. I think the charitable interpretation is that for most purposes distinguishing the two projects doesn’t matter much because, as I say, the most prominent expressivists combine them. But I might be wrong about that.
Thanks, Eric, that`s really helpful—I hope I don`t disappoint.
I take it you don`t just intend that they`re normative in the sense that any action, here like judging, is normative. I think there are two additional ways in which judging whether claims like (A/a – C/c) hold may be normative. First, given that (A/a – C/c) hold or do not, there is the (non-epistemic) question of whether we should believe them, assert them, or tattoo them on our knuckles. Second, there is also, as you mention in your response to Camil, the question of whether or not we endorse the legitimacy of some sort of discourse by engaging in it, even negatively. Another normative question. But that there is a connection between accepting (A/a-C/c) and some normative questions hardly shows that whether to accept them is (always) a fundamentally normative question in any sense other than the trivial one. What is necessary is to show that there is no other sense to be made of (A/a-C/c). This seems to be something you accept (is this right?) or accept a cousin of, given that you say
“The deflationist/expressivist approach helps us see that the question whether we should accept that there are moral facts cannot be addressed, even in part, by asking a prior question whether there are *actually* moral facts (just as we cannot ask whether we should accept that recreational slaughter is wrong by asking a prior question about its wrongness).“
But I’m calling shenanigans. Consider witch talk. Surely whether we should accept that he`s a witch is sometimes partially addressed by the question of whether or not he is actually a witch. Of course, maybe even though he is, actually, a witch, I shouldn`t believe it because (a) witch talk is not to be endorsed, (b) I`d be better off not believing him a witch, etc. Anyways, if this quoted bit is true, it’s very very surprising.
But maybe I still don’t understand the view? I should read your paper (which sounds really interesting.)
Hi Brad,
Sorry for being slow to respond to your post — I just returned from being on the road. (Ever been to Rumford, Maine? Not a whole lot up there!)
In response, I’m going to echo something I mentioned above in response to queries about how to spell out what it is for beliefs to track (non-contingent) elements of reality. What cases of veridical hallucination help us to see, I think, is that there has to be some sort of relation between a knower/knowledge states, on the one hand, and the facts known, on the other. (Arguably, it’s this conviction that drives the first horn of Sharon’s dilemma for normative realism.) The worry about Thick expressivism is that it’s difficult to see how there could be any such relation between states of moral knowledge and the moral facts known. The view rules out the possibility that moral facts can explain our states of moral knowledge. Third-factor views are a possibility, but I worry that they’ll be difficult to spell out in an Thick expressivist-friendly way.
Hi, Terence.
I just want to note that there is another concern implicit in your comments besides the Benacerraf-Field-Street epistemological concern. Previously you asked, “how could we get necessary facts/truths in mind such that we can…form de re/predicative beliefs about them?”, and you seem to be partly repeating that concern now. This problem sounds more like that of how terms like “good” determinately apply (given that goodness does not participate in causal interactions with us) than it does like that of how, given that our moral beliefs are (actually) true, we can explain their reliability. This problem is similar to Benacerraf’s *other* problem of “What Numbers Could Not Be” (and to Putnam’s of “Models and Reality”). I do not think that this problem can play the same dialectical role as the epistemological one.