Thanks to Brad Cokelet and the PEA Soup crew for the invitation to join an illustrious line-up!
Earlier this year my first book was published (Confusion of Tongues: A Theory of Normative Language, OUP). In the first part, I offer unifying semantic analyses for central, thin normative terms ‘good’, ‘ought’, and ‘reason’ (an “end-relational” theory). In the second part, I argue that when supplemented with a sensitivity to pragmatics, this theory solves many central problems of metaethics, including puzzles about practicality, categoricity, final value, and disagreement. A general theme is that metaethical puzzles largely result from philosophers’ confusion about our own language.
I’m very happy here to discuss any questions or objections readers might have. However, I thought I’d use the opportunity to focus particularly on meta-metaethical issues about philosophical method, as I’m currently trying to finish a paper for a volume on empirical approaches to metaethics (eds. Cuneo & Loeb), loosely based around a chapter I cut from the book at the last moment. I’m sticking my neck out here, because I don’t have broad expertise in metaphilosophy, so probably some or much of what I say is naive. But it seems an ideal topic for a blog discussion (honoring Sobel’s “plea for half-assedness”). I enthusiastically welcome any suggestions of work I should be reading or citing, including your own.
“The Empirical Armchair”
I argue that a particular kind of armchair, analytic method is at once (i) a viable, and (ii) an empirical approach to answering metaethical questions about the nature of normative properties. I also suggest it is (iii) uniquely apt, and (iv) the dominant actual method, despite what metaethicists may claim to be doing. (Hopefully that rattles a few chains).
The method I support seeks to identify the nature of normative properties (relations, etc.) by analyzing the conceptual meanings of the words we use to refer to them (like ‘good’, ‘ought’, ‘reason’, etc.), hence semantic/linguistic analysis. Because our concepts are often opaque to us in reflection, my method is to gather data about the various contexts in which we would and would not use particular words/sentences—as provided introspectively by our linguistic intuitions—and then reach an analysis of the concept abductively: what theory best explains this extension of use? Although this is done from the armchair I claim it is empirical, as it involves abductive inference from observational data causally connected with the target phenomenon. I also take it to be essentially the established practice in linguistic semantics.
My conception of the analytic method is in at least three ways more expansive than that of many philosophers in the past. First, it looks beyond narrowly moral uses, at the full range of ways the words are used. This includes more broadly normative or evaluative uses (e.g. ‘good’ in talk about good paper airplanes, jokes—in the tradition of Geach, Thomson). It also includes uses that aren’t normative at all, such as epistemic/predictive uses of ‘ought’, and purely explanatory uses of ‘reason’. Many philosophers dismiss these uses as involving homonyms; however, I argue (following linguists like Angelika Kratzer) that unifying accounts turn out to be highly promising. (‘Ought’ in terms of most possibilities, ‘reason’ in terms of explanation why, etc.)
Second, my method looks not simply for synonymy intuitions, but a full range of intuitions about sentential contexts in which these words can be used, and the general meanings of the words they combine with. For example, the way ‘good’ takes complementizers like ‘to φ’, ‘for s’, ‘for φ-ing’, etc., which points towards analysis of logical incompleteness (following Thomson, again), the way features of tense and aspect determine the availability of normative readings of ‘ought’/ ‘must’ conditionals (compare ‘Before I give you a license, you have to have filled out an application’ (normative) with ‘Since I gave you a license, you have to have filled out an application’ (epistemic)). Other results include a surprising analysis of ‘good for its own sake’ in terms of raising its own probability—which is supported by otherwise anomalous intuitions that ‘bad/ better/ best/ worse/ worst for its own sake’ are all peculiar (I’ve posted on this before, here), and identification of significant ambiguity in talk of “having reasons”—in virtue of the general vagueness of the possessive (e.g. ‘I have a master/ slave/ car/ pimple/ question.’)
The third feature is that the abduction is holistic: what counts as the best explanation is in part what is most compatible with our best theories in other domains—linguistics, psychology, epistemology, etc. The armchair philosopher should arm herself with as much of this knowledge as she can.
The second part of the project involves responding to multiple objections to this method; I’m presently counting eight. Some of these turn out not really to be challenges to the analytic method at all; this includes what I call (1) the “primitivist’s challenge”, (2) the “noncognitivist’s challenge”, (3) the “error theorist’s challenge”, (4) the “revisionist’s challenge”, and (5) the “pluralist’s [or perhaps, X-Phi] challenge”. (Hopefully the labels make it relatively clear what these are). The remaining challenges are genuine, but I believe the empirical evidence counts strongly against them. They are:
(6) The “skeptic’s challenge”: our intuitions aren’t sufficiently reliable (theoretical bias, etc.) to be good evidence of our concepts (e.g. Devitt?). I think the evidence lays this fear to rest. For example, empirical studies have found that linguists’ armchair results predicted surveys of the folk with 95% accuracy (Sprouse, Schutze, and Almeida), and philosophers have also been found highly reliable at predicting the results of allegedly surprising X-Phi experiments (Dunaway, Edmonds, and Manley, 2013). Anecdotally, I checked my surprising intuitions about the peculiarity of ‘it is bad/ better/ best/ worse/ worst for its own sake’ with a google usage search, with the following, striking confirmation: almost 5 million hits for ‘it is good for its own sake’, and only 2 hits, total, for all the other strings combined.
(7) The “synthetic realist’s challenge”. Perhaps the essential definitions of normative properties are synthetic rather than analytic? Perhaps, but all the evidence points the other way. (i) Unlike a variety of analytic hypotheses (primitivism, noncognitivism, logical incompleteness), this offers no explanation of the open question phenomena. (ii) Normative properties are evidently abstract/have a “narrow cosmological role” (since even things like ideas can be good), unlike the paradigms of things with synthetic definitions (like water), which causally impact the world in a wide variety of ways, and therefore are often picked out by their accidents. (iii) We demand more for semantic competence with normative words than a mere ability to refer; e.g. knowledge of supervenience, of practical significance, arguably the wrongness of torturing babies, etc. (iv) A little reflection on at least nonmoral uses of words like ‘good’ immediately suggests plausible analytic definitions.
(8) The “phenomenologist’s challenge”: why study normative properties via our language, rather than directly? (“To the things themselves!”) Answer: every inquiry needs to begin from some concept/ way of individuating its object. For some inquiries, it isn’t the quarry’s nature per se that escapes us, but which object is in question. I believe that metaethical inquiry largely begins from puzzlement about our own speech and thought: what do we mean by ‘good’, or ‘reasons’, etc.? In this case, normative language is like a treasure map: it’s our only clue to what properties we’re looking for. I think the evidence—and actual metaethical practice—overwhelmingly supports this diagnosis. (i) It explains the instinct to retire to the armchair (we have no such instinct when confronted with the question, “What is water?”). (ii) The mere fact that expressivism is a contender. (iii) Metaethicists don’t have any (other) uncontroversial ways of identifying the quarry. (iv) Ordinary folk have a competence with normative language that outstrips philosophers’ ability to explain how and why we use it the way we do. All in all, this is the profile for a quest for self-understanding.
I hope this isn’t too condensed to be understandable, and I look forward to your reactions…
Hi Steve –
Thanks for the thoughtful post! I have a question that might be about a side-issue, but I found the observation about “good for its own sake” vs. “bad/better for its own sake” quite interesting. Thinking about it, it seemed to me that even “good for its own sake” is an odd locution. (Normally, things are wanted/sought/done for their own sakes.) Then I remembered that one can “do good for its own sake,” which reminded me that one can also “do evil for its own sake.” Google’s ngram viewer confirms that the latter has a comparable rate of occurrence to the former. Does this fit your expectations? (And, could you say a little more about this fascinating observation anyway?)
Hi Tim!
Thanks for the question. To give a full explanation, I think I would end up just repeating my earlier blog post on the topic, which is here: http://peasoup.typepad.com/peasoup/2013/11/for-ss-sake.html. The fullest discussion is in my book, pp. 197-206.
But to answer your specific questions: it is common enough to talk about desiring things for their own sake, and this extends to frequent talk about desiring THE BAD for its own sake. Here ‘for its own sake’ is really modifying ‘desiring’. That’s why I searched on the longer strings, ‘It is good for its own sake’ and ‘It is bad for its own sake’–getting 4.6 million hits on the first, and only 1 hit on the second. So even if ‘good for its own sake’ is somehow odd, as you suggest, this hasn’t prevented widespread usage.
The central problem for my analysis here (‘p is good for its own sake’ means that p raises the probability that p) is that it is trivially true of everything. The pragmatic solution I alluded to is that the things we judge “good, for their own sakes” are, first of all, things we judge good simpliciter, which are those things that we judge good for saliently desired ends. So to say that p is good for its own sake is to communicate that p is good for/raises the probability of some desired end, and that desired end is p itself. (By contrast, to say that p is bad for its own sake would communicate that p lowers the probability of some desired end, and that desired end is p itself–which is impossible).
Let me emphasize my reason for raising this example here. I think it is good evidence that there is an implicit understanding of the (end-relational) nature of goodness in ordinary competence with normative words and concepts, even though this is opaque to us in philosophical reflection. I take it that even people who incline to absolutist metaethical views about the nature of goodness possess this reticence with regard to ‘bad for its own sake’, which shows that their metaethical theories don’t align with their own implicit concepts. And I think that’s a striking result.
Hi Steve,
Thanks for this intriguing post! I’m very sympathetic to much of what you say. I was just wondering if you could clarify whether the analytic method you’re defending is meant to be deployed in answering 1) metaethical questions, 2) first-order ethical questions or 3) both.
It seems plausible to me that 1) can be effectively answered via the analytic method you describe (i.e. reflecting on our linguistic intuitions involving words like good and ought and then homing in on the theory that best explains the observed patterns of usage). After all, if what we are interested in is “self understanding” (as you put it) — i.e. our moral practices and the concepts that figure into them – then investigating these concepts seems like a good way to acquire knowledge about these practices.
But if the analytic method you sketch is meant to address 2) as well, i.e. first-order questions about what to do, what’s good etc., then I wonder if the familiar worry about “losing normativity” isn’t liable to pop up. After all, even supposing we develop a good account about how words like “ought” or “good” are used, as well as the sorts of cases to which each of these words apply, this would only be an account of our actual usage. And so it remains possible to wonder whether this usage is *correct* — i.e. whether the way in which we use these words is the way that they actually should be used. Accordingly, one might worry that the analytic method you sketch does not provide to genuinely normative answers to first-order questions about what to do, what’s good, etc. At best, it seems only to provide a descriptive account of how we are actually inclined to answer these questions. (At least, that’s the worry.)
So does this mean your proposed analytic methodology is only meant to answer 1)? Or if it is meant to answer also 2), how would you answer the worry that normativity gets lost? Thanks so much for clarifying! Look forward to hearing what you think.
Hi Alex,
Mostly, I think it answers metaethical rather than normative questions. But that’s only because I think that ‘good’, ‘ought’ etc. are semantically contextualist/relativistic. If ‘good’ semantically expressed a determinate property, then semantic analysis would have direct normative results. (As Bentham and Mill seem to suggest). If I’m right, then semantics plus ordinary empirical knowledge tells us a lot about what we might call relativistic normative facts–e.g. that brushing your teeth is good for preserving your dental health–but doesn’t by itself settle moral disputes, which partly involve taking a stance on which ends to prefer. (Here I’m with the expressivists). However, once we fix on our moral ends, it does yield normative results directly.
As to how the words “ought” to be used: well, according to my own views, there’s a lot of different things one might mean by that. My theory commits me to distinguishing at least semantic, factual, and “moral” mistakes in using normative language. (Semantic mistakes involve using it in ways that violate its meaning; factual mistakes involve using it to assert falsehoods; “moral” mistakes get a quasi-expressivist analysis: they’re the uses we reject as expressing an intolerable set of preferences.)
Hi Steve,
Happy to see this discussion happening! I’m really interested in your pragmatics-based explanation of normative language’s practicality.
First, I’m curious how your approach answers worries with pragmatics-based explanation in general. We’re both familiar with the notion of the “pragmatic wastebasket” — the place we put communicative phenomena for which we lack a systematic semantic explanation. Semantics’ systematicity, with its attendant constraints from composition rules, syntax and psychology, makes it explanatory. But it is precisely because certain phenomena are not explicable within these constraints that we need to appeal to other non-linguistic information like speakers’ motives. What non-linguistic resources do you use to explain practicality and why use those resources in particular? How do those resources offer linguistic explanations?
More generally, I’m curious about your account of why words like ‘ought’ and ‘must’ are practical. If practicality isn’t located in semantic content, then how do assertions using the aforementioned words gain practicality as part of their meaning? I can’t see how Gricean maxims or souped-up Neo-Gricean versions would do the job.
Relatedly, how does your pragmatic account explain the connection between practical belief (supposing that practical judgments are beliefs) and motivation? If practicality is a pragmatic phenomenon, then the failure to be motivated by a practical belief isn’t a rational failure, as many cognitivists hold. For example, you can be competent with a language but incompetent with its pragmatic conventions. In that case, suppose you ask Mark where his keys are, and he pauses then responds “They might be in my office”. When you believe what Mark says but do not also form the belief that Mark does not know where his keys are, you are not irrational for failing to have the second belief. Similarly, if I say “you ought to dance, Steve” and you believe me, you are not irrational for failing to be motivated to dance, on your account. If not rationality or semantic content, what explains the observable and regular connection between practical belief and motivation?
Nathan, your last question seems relatively easy to answer on a Gricean or Neo-Gricean picture. Why is there a observable and regular connection between practical belief and motivation? Well, first of all, there isn`t such a connection. Unfortunately. If I had a nickel…
There is, maybe, a observable and semi-regular connection between practical assertion and motivation. And why is that? Well, typically the point of making a practical assertion is to either make clear where you stand, to advise others, or the like. Just like the point of saying that the restaurant is really expensive is typically to give a reason against going to the restaurant. And the point of an utterance is generally inferable from context.
When the point is the make clear where you stand, we generally expect that you endorse what you assert is wrong or ought to be done in the sense of having the proper motivational attitudes. If not, then why are you saying it? But not always (perhaps this connection is cancelable like a type of conversational implicature-I`ve suggested here and there that it is just like this. It might also be conventional, but I doubt it).
When the point is to advise, we generally expect that you advise relative to standards you accept and enforce on your own life (again, not always, and in ways entirely predictable.) Why? Your guess is as good as mine. But we do typically expect this.
Is this a rational failure? In a way. Systematic failure to understand then when I ask whether there is a convenient store which sells cigarettes nearby, the point of my asking is to find out something like
[where I can go get a pack RIGHT NOW]
is typically a kind of failure of rationality (unless you have no idea about how addiction works and the like). At least as far as I understand what a failure of rationality is. It is not irrational to not be motivated by sincere normative assertion because the connection between sincere normative assertion and motivation is not entirely robust. But when you systematically fail to see that the point of your utterance, as interpreted by ordinary speakers, generates an expectation that you will have the proper motivational states, then perhaps you exhibit a failure of rationality.
There may be an extensional problem here, but the outline of the story is, I reckon, clear, promising, and really attractive. Best of all, it slots explaining the motivational pull of practical assertion right back into a widely accepted and not-to-be-rejected-without-a-damn-good-reason research program.
Hi Nathan,
Thanks! My full answers to (as it happens) both these questions are in Ch. 5, “Pragmatics and Practicality”.
I try to put my pragmatic claims on a sound footing by identifying a single basic principle of pragmatics, and deriving all my results from that. I write, “The basic character of pragmatic reasoning is this: what can be inferred from the fact that the speaker believed that uttering THAT particular sentence was best for promoting her conversational ends in the context?” (p. 123)
Regarding practicality: according to me, normative uses of ‘must’ and ‘ought’ (etc.) are end-relational; e.g. to say that S “must” do A is to say that if some particular end is to be realized, then it is necessary that S does A. When we don’t explicitly relativize such claims (as in moral discourse), the end is the saliently preferred end. For purposes of illustration, lets suppose this is the agent’s/subject’s preferred end. So if I tell you that you must do A, I’m telling you that doing A is necessary for realizing your preferred end e. If you believe me, your motivation toward e is what motivates you to do A. I don’t think there’s any mystery about how such beliefs are practical.
On the connection between pragmatics and rationality: I want to ask, what is it that I come to believe, when I accept that I ought to dance? If my semantics is correct, then this only has a propositional content if it is relativized to some end. If I understand this end to be my preferred end, then I would be instrumentally irrational if I failed to have motivation to dance. (FWIW, I don’t believe in instrumental irrationality, or indeed in irrationality in general, but that’s a whole other can of worms). However, you’re presenting a case where I’m pragmatically incompetent, so let’s suppose that I fail to understand that you mean my preferred end. (Perhaps, for example, I assume you to mean that I ought to dance in order to promote moral ends). In that case, I think you’ll agree that there is no expectation of rational motivation, and that the case is just like the ‘might’ case.
Jack, thanks–I replied to Nathan before seeing your post. I basically agree with everything you say. (Right down to the cancellable conversational implicature).
Thanks for this very nice piece! I’ve really enjoyed the individual papers coming out of this larger vision, and it’s great to see that you are now taking on these broader methodological issues.
I was just wondering if you could say a little bit more about the pluralist’s objection. An obvious view about, say, the study of natural language deontic modals is that it should involve a mix of armchair and experimental methods. (This is certainly what one finds in contemporary semantics journals.) Are you disagreeing with that view, or do you see it as compatible with what you propose here?
Hi Josh,
It’s a bit scary to be taking on the methodological issues, to be honest with you, given my ignorance of so much of the relevant literature, but I figure that any self-respecting philosopher should be willing to reflect on and defend his or her methods! And since my methods seem to be at odds with what metaethicists have been saying for the past 50 odd years, I perceive something of an obligation. (I’m very appreciative, BTW, of your interest and encouragement.)
I absolutely agree that natural language should be studied through both armchair and more experimental methods. I think I’ve learned valuable things from both comparative linguistics and etymology, for example. So now you have me wondering if I overstated my case. But I would argue that there’s no need for the metaethicist to do the experimental or fieldwork herself.
The “pluralist’s objection” I had in mind is, more particularly, that armchair methods should be forsaken because our concepts/meanings are idiosyncratic or parochial. To that I respond, first, that the empirical evidence in this case says otherwise, and second, that even if it were true, it would be largely irrelevant to the viability of conceptual analysis as a way of answering our own metaethical questions.
But I’m still feeling my way around the various subtle issues here, so I value hearing everyone’s opinions.
Hi everyone,
A super quick (mostly clarificatory) follow-up to Nathan:
Nathan writes, “Semantics‘ systematicity, with its attendant constraints from composition rules, syntax and psychology, makes it explanatory. But it is precisely because certain phenomena are not explicable within these constraints that we need to appeal to other non-linguistic information like speakers’ motives.
I feel like I’m missing something here. “Motives” sound psychological to me, and Nathan says that semantics has built into it constraints concerning psychology. What is it about “non-linguistic information like speakers’ motives” that makes such information “non-linguistic” (non-semantic?) if we’re understanding semantics as including constraints concerning psychology?
Nathan also writes, “Relatedly, how does [Steve’s] pragmatic account explain the connection between practical belief (supposing that practical judgments are beliefs) and motivation?” Not to preempt Steve, but I would have thought that what explains the connection between “practical belief” and motivation for Steve is that practical beliefs just are ordinary descriptive beliefs held in the presence of motivational states (desires).
Looks like my comment didn’t get in quick enough!
Hi Nick,
Well, I scooped you on the point about practicality, but I failed to pick up on the semantics/psychology issue. Thanks!
Hi Steve,
This is fun stuff, and I’m very interested in your take. I want to gather some of your key claims about how empirical your methods are.
1. A referring term T has a conceptual meaning.
2. Introspectively accessible intuitions about how one would use T in various contexts comprise data relevant to which conceptual meaning T has.
a) Introspection on intuition counts as observational data.
b) This data is causally connected to the referent of T.
3. We abduct from the data to form a best theory of T’s conceptual meaning.
I see that these key claims identify a method that has many analogies with prototypical ways of going about empirical investigation (abduction from observational data). But I wonder what the evidence is for the key claims themselves, especially claims (1) and (2). If they are the conclusions of abduction from data, or more generally best explanations of explanada, I think empiricists would be more comfortable with the picture, though I imagine some will vigorously dispute whether these key claims are supported by the data. I can imagine many balking at the idea that the data supports the hypothesis that the referring terms of interest have conceptual meanings, for example. I’m sympathetic with you here, but the task seems to require adducing all that good (empirical) evidence for conceptual meanings and how intuitions are connected to them. These empirically supported hypotheses in hand, we can then proceed with confidence with the method you identify: abducting from intuitions to conclusions about conceptual meanings. Is that the picture?
Separately, maybe you could explain this to me:
“(i) Unlike a variety of analytic hypotheses (primitivism, noncognitivism, logical incompleteness), this [synthetic realism] offers no explanation of the open question phenomena.” They have some explanation, don’t they? Namely, Moorean open feels come from the absence of analytic definitions of normative terms in non-normative terms. Maybe not the best explanation, but that is some explanation. Am I missing your point?
Thanks, Matt!
These are all meant to be empirical claims, and I think there is evidence for them all. Some but not all of this evidence is armchair evidence from intuition, so I’m inclined to think that we can be justified in proceeding with our armchair analysis in confidence, on the basis of solely armchair considerations. However, I’m not strongly committed to that, as I think there’s plenty of non-armchair support as well. See Brian Talbot’s work, for example.
Re your (1): I’m not saying that every referring term has a conceptual meaning. I don’t think that’s true of names, for example. But I do think it’s true of central normative terms like ‘good’ and ‘ought’. One piece of evidence is that we demand quite a bit before we allow that someone is competent with these words, unlike names: e.g. knowledge of supervenience, practical significance (“sure it’s a reason for me to do A, but what’s that to me?”). There are a lot of issues here, of course, about semantic externalism and the like, which I’m happy to discuss if you push me. (I might have to post a follow-up about some of these issues).
I’m understanding synthetic realism as the claim that the essential definitions of normative properties are synthetic. This does explain why the essential definitions don’t produce closed questions. But it doesn’t explain why there wouldn’t be any closed questions at all. Synthetic realism doesn’t by itself entail that we don’t use concepts to refer to those properties. (For example, perhaps we use(d) the extrinsic and non-essential concept “the actual intoxicating ingredient of wine” to refer to (ethyl) alcohol). The lack of closed questions in relation to normative terms therefore is still in need of explanation. (If I recall, this point is made nicely by Charles Pigden in his paper “Identifying Goodness” (AJP 2011)).
Hi Steve,
I’ve really enjoyed reading your papers on the end-relational view. I’m afraid I haven’t read the book yet (because I’m lazy!), so I apologize if these questions/complaints are dealt with there, but I want to try to offer some pushback on behalf of what you’re calling the phenomenological approach (which I suspect is actually several rival approaches). I guess I find your semantic picture really compelling. But I don’t know how to apply it to what I’ve always found to be the most interesting parts of metaethics, which I’ve taken to be motivated by the experience of decision-making, of exercising self-control, etc. (Think Frankfurt, Watson, Velleman)
I do experience on occasion something that feels like deliberating about what ends to adopt. I agree that in a lot of cases, my ends are set, and deliberation is largely a matter of figuring out the most efficient way of obtaining them (and maybe glancing around to make sure I won’t accidentally frustrate another end in the process). But in some cases it seems like I don’t have set preferences, and part of the point of deliberation is deciding them. This is most obvious in moral cases: when I try to figure out whether to give money to charity it (trust me: I have no antecedent preference for giving my money away). But also decisions to make more time to read novels. It doesn’t seem like I have any antecedent preference for reading X novels a year rather than Y. (And it doesn’t seem like what was keeping me from reading X novels was some ignorance about instrumental relations.) So I guess I’m wondering, if I say to myself “I wonder if I ought to read more stuff that isn’t philosophy” and conclude yes, and so read more, what are the ends in relation to which the ‘ought’ should be understood? I don’t think it could be my existing preferences– unless you want to adopt a debunking story about what seems like non-instrumental deliberation. But then what?
More generally, it seems like some metaethical theories have appealed to non-semantic considerations, such as: general features reasoning would have to meet to count as practical reasoning, the nature of desires, plausible constraints of rational coherence on our attitudes, and so on, to justify certain theories of practical reasons or value or whatever. I would argue that these kinds of psychological considerations have actually been far more important in motivating expressivism than semantic issues– and the evidence is that expressivism has struggled mightily to adequately deal with the latter, and yet people continue to find it compelling.
To take this back to what I said in the beginning, I find the semantics for “ought” that I’ve read in your papers compelling. But let’s say some version of Kant’s argument for the categorical imperative is compelling. Doesn’t that tell us that there’s one contextually salient use of “ought” that is philosophically more interesting than the others, given the way it answers (for example) to necessary aspects of rational decision-making? Or another way of putting this, I don’t see how your semantics is in competition with a lot of other metaethical views. Can’t I accept it, and just keep thinking most of the rest of what I thought before. For another example, couldn’t I accept the end-relational view, but be a Moorean non-naturalist who thinks that in some contexts the relevant ends are determined by which properties happen to have Moorean halos circling them? All of which is to say, I guess, why the confidence that semantic analysis does everything?
Anyway, I apologize if there’s been too much rambling, but maybe you could say a bit more about how you’re thinking about this. And obviously I need to read your book.
Thanks Derek, these are astute objections, that make me think I’ve overstated my case here.
One issue is what counts as “metaethics”. You’re pointing to places action theory and normative ethics touch on metaethics, and I’m not inclined to say that either of these areas of philosophy is best pursued via semantic analysis. (Though I think it bears usefully on them, in places). I was thinking, probably too narrowly, of metaphysical questions like “what is value/goodness/the relation of being a reason?” But I should acknowledge that the purview of metaethics is much broader than that.
However, the more serious challenge you’re raising is that even these narrowly metaphysical questions can’t be addressed just by semantic analysis, because they can’t be disentangled from the psychological/action theoretic/normative questions. I concede the point. It’s supposed to be captured by the third feature of the “enhanced” analytic method I described: that it’s holistic, looking for the best explanation within the context of our best theories in other domains (including moral psychology and action theory). I would still maintain, however, that semantic analysis makes up the largest part of the most appropriate method of answering the narrowly metaphysical questions.
The end-relational semantics do not, as you say, by themselves refute moral/metaethical rationalism, or nonnaturalism (Nick Laskowski has a paper in Res Philosophica about this). My strategy in the book is to argue, first, that the theory takes away the main positive argument for those views and against Humean/subjectivist views (i.e. the argument that the Humean subjectivist is unable to explain our practices of normative speech and judgment), and second, that considerations of simplicity and conservatism then count against those views (rationally required ends? nonnatural properties?) You COULD combine an end-relational semantics with a nonnaturalist metaphysics (in the book, I discuss this in relation to a nonnatural property of importance, possessed by some ends [pp. 252-3]). But once you have an end-relational semantics, you don’t need a nonnaturalist metaphysics. Or so I claim.
But I acknowledge (e.g. p. 251) that this is hostage to the kinds of action-theoretic and psychological concerns you’re pushing. Originally, the book was supposed to have three parts: semantics, pragmatics, moral psychology–but this turned out to be too ambitious. So my Humean conclusions are provisional, and I hope to get back to working on issues in moral psychology soon.
Finally, deliberation about ends. I agree that this is a plausible place to push back against my Humean stance. In the book, I throw a few things at the problem. Perhaps the most promising is “what ought I to do, in order to realize the end I would most prefer if I were fully informed?” (Note that the end-place can be bound by a description, so there doesn’t have to be a particular end you have in mind.) The key question is: what motivates deliberation about ends? (Of course, this isn’t a semantic question). The absolutist will say something like “a desire to pursue what’s right/best”. But I don’t find that very plausible. Couldn’t we also deliberate about whether to pursue what’s right/best, if that were just some objective property some ends have?
In response to Derek’s question about deliberating about what ends to adopt, Steve, you suggest appealing to ends one would adopt with full information. However, I wonder whether something like the following might provide a more straightforward answer to Derek’s objection.
In particular, when one deliberates about what ends to adopt and then forms a judgment like “I ought to adopt the end of reading more non-philosophy books,” why not say that the end relative to which this judgment is made is some subset of the ends to which you already adhere. Thus, for example, you might think that reading more philosophy books is useful as a means to accomplishing my antecedent ends of acquiring a broad knowledge base, becoming a more relaxed person, experiencing more pleasure, and so on. Call these ends “E1-E3,” while the end of reading more non-philosophy books is “E4.” The judgment that you ought to adopt the end of reading more philosophy books, then, would amount to the judgment that adopting E4 would promote the satisfaction of E1-E3. (Or something along those lines.) The trick is to treat adopting a particular end as just another action one might perform and then run the end-relative analysis of ought judgments as per usual. Wouldn’t this do the trick? (I, too, have not yet read your book and apologize for all the way in which I am surely botching the details of your view.)
Thanks!
Alex, I do think something broadly coherentist like this is another option. However, there are some complications. My semantics is end-relative, not ends-relative, and I don’t think it’s plausible that every case of deliberation about ends is framed by a question like “What ends ought I to pursue in order to achieve the conjunction of E1-E3?” This is why I currently prefer a bound variable in the end argument. (I devote a whole chapter, Ch. 6, to issues arising from multiple ends).
I’ll add, though, that “reading more non-philosophy books” doesn’t strike me as a plausible candidate for a FINAL end, so I take it there are cases that present a greater challenge. E.g. Sartre’s dilemma, between family and country.
Thanks for the response. That sounds really good.
However: “I’ll add, though, that “reading more non-philosophy books” doesn’t strike me as a plausible candidate for a FINAL end…”
But reading novels instrumentally is barbaric.
I’d cop to being a barbarian, but I’ve barely read a novel in the last 15 years!
(I think I owe some fuller responses to some of the questions; I’ll try to do that tomorrow.)
…So I never got around to expanding on my responses. Perhaps this discussion has come to an end, but (especially in light of the holiday break), I’ll continue to check periodically as long as the comments thread remains open. So feel free to jump in anytime.