+ All Categories
Home > Documents > Practical Realism?

Practical Realism?

Date post: 06-Nov-2016
Category:
Upload: john-hawthorne
View: 215 times
Download: 0 times
Share this document with a friend
10
Philosophy and Phenomenological Research Vol. LXIV, No. 1, January 2002 Practical Realism?’ JOHN HAWTHORNE Rutgers University In ‘Normative and Recognitional Concepts’, Allan Gibbard attempts to combine a sort of naturalistic moral realism with some of the main threads of quasi-realism. While his piece is certainly rich and suggestive, I found it unpersuasive at almost every key step. Below, I detail six areas of puzzle- ment. 1. Realism and Quasi-Realism ‘Normative and Recognitional Concepts’ is strikingly realist in its orien- tation. To make this vivid, let me contrast Gibbard’s current view with that of his earlier self. In Wise Choices, Apt Feelings: he defended a moral anti- realism that embraced, inter alia, the following two key ideas: (1) Non-Descriptivism: Moral claims and thoughts are not, fundamen- tally, attempts to describe the world. “Normative talk is part of nature, but it does not describe nature. In particu- lar, a person who calls something rational or irrational is not describing his own state of mind; he is expressing it. “3 On this view, surface syntax is not a reliable guide to whether an utterance purports to describe: Superficially declarative syntax masks the constrast between talk that is in the business of describing and talk that is in the business of doing something else. (2) Non-Factualism: Moral claims and thoughts do not, strictly speak- ing, express facts. “In my own picture, all strict facts will be naturalistic.. .. Apparent normative facts will come out, strictly, as no real facts at all”4 I * Thanks to Tamar Gendler and Ted Sider for helpful comments Cambridge, Mass: Harvard University Press, 1990. Wise Choices. Apt Feelings, pp. 7-8. Wise Choices, Apt Feelings. p. 23. SYMPOSIUM 169
Transcript

Philosophy and Phenomenological Research Vol. LXIV, No. 1, January 2002

Practical Realism?’

JOHN HAWTHORNE

Rutgers University

In ‘Normative and Recognitional Concepts’, Allan Gibbard attempts to combine a sort of naturalistic moral realism with some of the main threads of quasi-realism. While his piece is certainly rich and suggestive, I found it unpersuasive at almost every key step. Below, I detail six areas of puzzle- ment.

1. Realism and Quasi-Realism ‘Normative and Recognitional Concepts’ is strikingly realist in its orien-

tation. To make this vivid, let me contrast Gibbard’s current view with that of his earlier self. In Wise Choices, Apt Feelings: he defended a moral anti- realism that embraced, inter alia, the following two key ideas:

(1) Non-Descriptivism: Moral claims and thoughts are not, fundamen- tally, attempts to describe the world.

“Normative talk is part of nature, but it does not describe nature. In particu- lar, a person who calls something rational or irrational is not describing his own state of mind; he is expressing it. “3

On this view, surface syntax is not a reliable guide to whether an utterance purports to describe: Superficially declarative syntax masks the constrast between talk that is in the business of describing and talk that is in the business of doing something else.

(2) Non-Factualism: Moral claims and thoughts do not, strictly speak- ing, express facts.

“In my own picture, all strict facts will be naturalistic.. . . Apparent normative facts will come out, strictly, as no real facts at all”4

I

* Thanks to Tamar Gendler and Ted Sider for helpful comments Cambridge, Mass: Harvard University Press, 1990. Wise Choices. Apt Feelings, pp. 7-8. Wise Choices, Apt Feelings. p. 23.

SYMPOSIUM 169

The Occurrence of ‘strictly’ is important of course. A sensible moral anti- realism will allow for the use of the term ‘true’ within moral discourse (such as the use of ‘That’s true’ to express agreement). The strategy is to pay lip service to realism without really being a realist. The moral anti-realist permits himself to say ‘It’s a fact that murder is wrong’ and ‘That’s true’ in moral argument, but when doing serious meta-ethics, she will assert “Strictly speaking, there are no moral facts”.

Are (1) and (2) orthogonal to each other? Well, we can certainly imagine embracing (2) while denying (1) for some fragment of discourse. When the Cretan says “All Cretans are liars”, he is purporting to describe the world even though he fails to assert anything that is true or false. Perhaps meta- physical, religious or moral discourse has a similar status: it does not get to be either true nor false even though it purports to describe the world. ‘Norma- tive and Recognitional Concepts’, meanwhile, appears to embrace (1) while denying (2).

Gibbard’s own presentation is somewhat distracting, covering up the fact that his new view is a substantial retraction of his old one. The contents of the paper are far from realizing Simon Blackburn’s quasi-realist program,’ which embraces both (1) and (2). For the new view does not merely pay lip service to the idea that moral discourse expresses facts against the background of a whispered “But it doesn’t really, strictly speaking, get to be true or false.” On the contrary, the new view is that moral predicates-such as ‘the thing to do’-get to, really and truly, pick out natural properties. When Gibbard says that ‘the thing to do’ picks out a natural property, he does not warn us against taking that claim at face valuea6 What is distinctive about his account is his story about how normative concepts pick out natural proper- ties.

That brings us to (1). Non-Descriptivism remains intact. Moral predicates, while picking out natural properties, are not fundamentally vehicles of description. The whole point of the new account is that there is more to being a descriptive concept than referring to a bona fide property in the world. Within the class of genuinely referential predicatedconcepts, we can still distinguish those whose canonical functional role is that of describing versus those whose canonical functional role is something else. ‘The thing to do’, it turns out, falls on the “something else” side of the fence. Here’s how I under- stand the contrast. The point of descriptive concepts within our cognitive architecture is either to track certain properties that are directly available in perception or else to provide theoretical, and especially causal, explanations

See, for example, ‘Evaluations, Projections and Quasi-Realism,’ in Spreading rhe Word, Oxford University Press, 1984 and Essays in Quasi-Realism, Oxford University Ress, 1995. If he still adheres to (2). he now does this so quietly that we cannot discern that he does so.

170 JOHN HAWTHORNE

of what we encounter in perception. On this score moral concepts are not descriptive. Their point is neither to describe what is directly available nor to explain; rather, their point is to direct action. Hence Gibbard’s practical real- ism: While moral concepts pick out properties in the world, the point of having them is as a vehicle for settling on actions.

In 92 below, I raise some concerns about Gibbard’s way of motivating non-descriptivism. 93 addresses the cogency of his argument for naturalism. 94 questions the recognitional constraint that he imposes on a theory of the good life. 55 questions his account of how logical connectives interact with moral predicates. 96 raises general concerns about the kind of naturalistic realism that Gibbard now endorses.

2. What to Do The backbone of Gibbard’s case for non-descriptivism in this paper is the purported equivalence between the question ‘What ought I to do?’ and the question ‘What to do?. I doubt whether this link can do the philosophical work that Gibbard asks of it.

What is certainly correct is the linguistic point that we frequently use infinitive constructions to express normative contents: “The move to make is to fork the rook and queen” “One way to help is to give him some money” obviously have normative content. Gibbard wishes to go further: he wants to connect ‘what to do’ with ‘what I ought to do’ in order to underscore the role of ‘ought to do’ as an expression of decision. Presumably the idea is that since ‘what to do’ conclusions express decisions and ‘ought to do’ conclu- sions are equivalent to ‘what to do’ conclusions, then ‘ought to do’ conclu- sions express decisions too. If this is the line of thought, it is unconvincing. Three points are relevant.

First, normative-expressing infinitives often express what is permissi- ble-what one can do-as opposed to what is obligatory-what one ought to do. (Witness the ‘one way to help’ example above.) So ‘what to do’ is ambiguous as between permission and obligation: and insofar as conclusions about what to do express the former, they are quite clearly not decisions to act.

Second, when I come to a conclusion about what someone else is to do, I do not, plausibly, express a decision: I straightforwardly express a belief. (‘The thing for Jim to do is to get out of there fast.’) So how can appeal to ‘what to do’ constructions tell against descriptivism?

Third, we can distinguish, it seems, between knowing what to do and deciding what to do: The akratic knows what to do and yet decides not to do it. So ‘what to do’ talk does not invariably express a decision to act. Of course, if one has no weakness of will then, insofar as one knows what to do, one will decide to do it. But that point could have been made just as well

SYMPOSIUM 171

without invoking the ‘what to do’ construction: Someone who has no weak- ness of will and knows that he ought to do x, decides to do x.

The question remains: what philosophical payoff is there in focussing on the ‘what to do’ construction?

3. Natural Properties and Goodness Gibbard offers an argument for the thesis that moral properties are constituted by natural properties. The argument assumes that which it is meant to prove.

Which property is the property of being the thing to do? Gibbard tells us that “to say what natural property this is would be to offer an answer, in complete and fundamental terms, to the question of how to live”.’ So the idea is that there is a true answer concerning how to live and that the natural prop- erty picked out by “what to do” can be read off from such an answer. The relevant answer will take the form of a universal plan, “a plan for what to do in any possible circumstance”:

A universal plan can take the infinite form, in situation S1 do the act with natural property PI, in situation Sz do the act with natural property P2, and so on.

He goes on:

From this, we can construct the grand property, having PI in S1, P2 in S2, etc. Call this prop- erty P *; the plan is, then, in any possible situation, to do the act with this grand property P*.. . . And this properly is constructed, finitely or infinitely, out of natural properties.*

I have two main worries. (i) Suppose I were to set about forming a universal plan, which c o v e f e d

any possible situation. Possible situations, not surprisingly, are plentiful. Some of them involve some humdrum recombination of the atoms in the void. Some of them do not. After all, there is a possible situation in which I see a ghost. And what should I do if I see a ghost? A good question for a universal planner. But the answer to this question will point beyond any property constructable finitely, or infinitely out of nurural properties. Gibbard thus assumes without argument that the universal planner can simply ignore questions of what to do in possible scenarios that instantiate non-natural properties.

(ii) Suppose goodness did not supervene on the natural world, that it was a sort of halo that could accompany or be absent from some physical totality. If I were a universal planner who did not think goodness supervenes on a non- normative groundfloor then I would be far from able to formulate a universal plan in terms of natural properties. Give me some naturalistic description of a world N and I will not be able to answer “What to do” questions until you

’ P. 158. P. 160.

172 JOHN HAWTHORNE

tell me how the normative properties are distributed. If N were consistent with two different distributions of haloes, so to speak, then I could hardly say what the thing to do was unless the description also included the distribution of normative properties. I am not, of course, proposing that we reject a global supervenience thesis in ethics according to which duplication of the naturalistic facts ensures duplicate answers to ‘What to do’ questions? I merely wish to point out that a global supervenience thesis is secured by Gibbard’s account only because he assumes from the outset that the morally omniscient universal planner can say what the thing to do is in any possible situation given only a non-normative description of that situation, which is to assume at the outset that two situations cannot be alike non-normatively and differ normatively. Global supervenience is, surely, the main import of the idea that moral properties are constituted by non-moral ones. Insofar as Gibbard believes himself to have offered an argument for the constitution thesis, he overlooks the fact that he assumes from the outset much of what he is supposed to be proving.

4. Recognitional Concepts and Plans for Life Gibbard imposes an untenable constraint on moral planning. Plans for life must be formulated, he tells us, in terms of recognitional concepts:

The building blocks of a contingency plan must be concepts that are recognitional, concepts in terms of which we can recognize the circumstance we’re in, and in terms of which we can settle what to do and so do it.”

He concludes that the property corresponding to “being the thing to do” must be composable from recognitionally grounded concepts. If the universal plan were not so framed, then it could not be implemented; and it is a requirement on a plan for life than one can implement it.

It is, however, no reasonable requirement on a plan for life that one be able to know for sure that one is implementing it. I can’t know for sure that I am not a brain in a vat. Yet I can still have it as part of my plan for life that if there is a drowning person in front of me, then, other things being equal, I’ll try to help. This point generalizes. “Don’t step on living creatures” can be part of my game plan even if there are cases where I cannot tell by obser- vation whether some creature on the road is living or dead. My plan isn’t to avoid stepping on things that appear to be living or that I judge to be living without hesitation. My plan is to avoid stepping on living things. I rely on my perceptual evidence in order to implement that plan as best I can. Of

This formulation needs a bit of refining. See the discussions on logical supervenience in David J. Chalmers, The Conscious Mind: In Search of a Fundamental Theory (Oxford University Press, 19%). Those subtleties need not concern us here.

l o P. 162.

SYMPOSIUM 173

course, if I could never even acquire evidence as to whether a thing was living or dead then it would be a deeply unimplementable game plan. But that is not the case here. Gibbard seems to be requiring that, ideally, a plan for life is such that in any situation I can recognize whether a certain basic action is the thing to do. That seems to require that plans for life avoid any concept for which I might in some perceptual situation be properly hesitant about whether it applies or not. A tall order indeed.” Isn’t it enough if I can from some reasonable hypothesis about what is the thing to do on the basis of the evidence available to me? Isn’t enough if I can know in a large range of cases-the cases where it is clear to me that the antecedent of a conditional plan is satisfied-what the thing to do is?

We can in some reasonable sense make an informed judgment about whether a fragment of some scientific theory is true even if that theory isn’t couched in recognitional terms. Why can’t I also in some reasonable sense implement a plan even if that theory isn’t couched in recognitional terms? The requirement that practical wisdom cleaves to the recognitionally available even though theoretical wisdom does not is undermotivated.

5. Logical Connectives and Moral Discourse Gibbard’s realism is hard to square with his felt need to provide an anti-realist semantics for logically complex constructions containing moral predicates. Let me explain.

Moral discourse is replete with logical connectives. This enables us to engage in deductively valid reasoning in the moral sphere and to recognize logical truths that contain normative language. This also enables us to embed declarativemoral sentences into more complex sentences in such a way that affirming the latter is not tantamount to either affirming or denying the former.

These phenomena make trouble for the more flat-footed versions of moral anti-realism. Suppose for example, that one thought (a) that utterances of the form “That’s good” are never true or false and that (b) one sufficiently under- stands “That’s good” once is able to use the latter as a device for expressing a pro-attitude. How can such a theory explain the status of

“If that’s good then that’s good”

‘ I In ‘Cognitive homelessness’ Journal oJPhilosophy 93, 1996, Timothy Williamson argues that it is an illusion to think that any interesting condition is luminous, where a condition is luminous just in case one is always in a position to discern whether it obtains or not. The key thought is that, as we move towards borderline cases of a concept, we will be hesitant about whether to apply it or not and hence will not know whether or not it is applicable. (This is true even of concepts such as looks red. So not even phenomenal conditions are luminous.) Gibbard seems to want a universal plan that is formulated in terms of luminous conditions. If Williamson is right, then what he wants is altogether unattainable.

174 JOHN HAWTHORNE

as a logical truth, since such a status would seem to require that the antece- dent and consequent are themselves capable of taking on a truth value (contra (b))? And how can such a theory explain the validity of the inference from

He’s not tall

to

He’s not both good and tall

since validity requires that whenever the premise is true the conclusion is true too?

Moreover, how could such a theory even explain how we can manage to understand ‘That’s good” as it occurs in an antecedent of a conditional, since the sketch of understanding under (b) above does not appear to facilitate such a use?

Gibbard is very concerned to deal with these problems. What should be noted, though, is that they are less acute in the first place for the practical realist. After all, these problems arose, in good measure, thanks to the moral anti-realist’s insistence that moral claims lack a truth value. This puts strain on such concepts as logical consequence and logical truth in connection with moral discourse. But that problem does not arise for Gibbard, since on his account, moral sentences do get to be made true or false by the distribution of natural properties.

The residual problem for Gibbard is to provide a plausible answer to the question “What is it to grasp the content of moral predicates?” We all want an account of understanding according to which we can understand moral predi- cates as they occur in logically complex sentences. Gibbard recognizes that such simple-minded accounts of understanding as

(c) One understands “x is the thing to do” once one is able to use that construction as a device for expressing a settled decision to do x

will not explain one’s understanding of the Occmnce of ‘x is the thing to do’ in logically complex sentences. In short, Gibbard want to provide an account of what it is to grasp “x is the thing to do” that supplements a story with the shape of (c) with a story about how we are able to understand embeddings of moral sentences within logically complex constructions. The story that he offers leans on the pair of attitudes “accepting” and “rejecting” , analysing the contribution of logical particles to complex formulae in terms of what attitudes are precluded by the complex formulate. Thus the meaning of

A or B

SYMPOSIUM 175

is constituted by that fact that acceptance of the above precludes rejecting both ‘A’ and ‘B’. Of course, “precludes” is normative. It is possible to accept ‘A or B’ and to reject both ‘A’ and ‘B’. The point is that one who understands ‘or’ would then, by the lights of his own understanding, have done something defective.

But what would be wrong with a more standard account of our grasp of the logical particles in terms of rationally compelling inference rules? Consider, for example, an account according to which we grasp the meaning of ‘and’ by mastering the introduction and elimination rules appropriate to conjunction.1Z On this story the meaning (or Fregean sense) of a logical connective is constituted by the canonical introduction and elimination rules (and, perhaps, transformation rules) in which it is caught up.

Gibbard does not justify his departure from more traditional approaches to the meaning of logical connectives. Worse still, I can’t see that his own style of analysis has much hope of providing an adequate account of their meaning. “It is irrational to accept I”’ and “Not P” both preclude accepting P. But it is hardly right to think that “Not” can be analyzed as “It is irrational to accept”. That means that the content of ‘Not’ cannot be given in terms of what it precludes accepting. One reason for the non-equivalence is that not accepting is hardly tantamount to rejecting. This point infects Gibbard’s discussion at various places. Consider for example his account of reasoning that proceeds by disjunctive syllogism:

With the first I rule out this: coming to reject A and to reject B. With the second I reject A. The combined effect, then, is to rule out rejecting B. It would therefore be inconsistent of me to accept the premises and reject the conclusion: I would be doing what I had ruled out doing.13

But this hardly explains why the conclusion is rationally compelling. For all Gibbard has said there would be nothing at all wrong with accepting the premises and trenchantly remaining agnostic about the conclusion, neither accepting nor rejecting it. Relatedly, ruling out an attitude that rules out rejecting P hardly seems tantamount to accepting P, since required agnosti- cism would be sufficient grounds for the former. So if one is classically inclined, one will not like the account of double negation that issues from Gibbard’s picture.

In sum, one is left wondering what would be wrong, even from Gibbard’s point of view, with a fully realist semantics which associates naturalistic properties and states of affairs with moral predicates and atomic moral sentences respectively and then assigns truth functions to the logical particles in accordance with the dictates of classical logic. One could supplement this

See, for example, Michael Dununett, The Logical Basis of Metaphysics, Harvard University Ress, 1991 and Christopher Peacocke, A Study of Concepts, MIT Press, 1992.

13 P. 155.

176 JOHN HAWTHORNE

account of reference with an account of the sense of moral concepts (making special mention of the relation of such concepts to action) and of logical particles (making special mention of canonical introduction and elimination rules). Given Gibbard’s denial of non-factualism, it is just not clear what prompts the eschewal of this more standard framework.

6. Realism and Semantic Mystery

Let me end by stepping back from the details of Gibbard’s argumentation in order to raise some general concerns about the kind of realism on offer. Gibbard wants to combine realism with naturalism: the property of being good exists all right but it comes to nothing more than some very compli- cated combination of natural properties, a combination that will presumably appear rather gerrymandered from the point of view of natural science. David Lewis uses the term ‘naturalness’ to describe the dew to which a property marks out a joint in nature.14 Extremely complicated combinations of funda- mental properties are, prima facie, rather unnatural. They have a certain natu- ralistic pedigree all right-after all, they are constructed from eminently respectable properties. But plenty of gruesome proper tie^'^ have that pedigree. That pedigree is thus hardly a mark of naturalness. Consider now another element of Lewis’ picture of things: the more natural a property is, the more eligible it is as a referent of language and thought. Insofar as a property is gruesome then, other things being equal, the ideal interpreter will not assign that property as the content of language and thought. In particular, the avail- ability of a more natural property that equally well fits other constraints on interpretation for some predicate will discount gruesome properties from seri- ous consideration when it comes to devising a semantics for that predicate. And if reference candidates are all equally ineligible, then reference is to that extent indeterminate.

Any moral realist who buys into the picture so far faces a serious worry: How is it that moral predicates such as ‘being good’ or ‘being just’ manage to lock on to the particular properties that they do? The properties in question appear to be rather gerrymandered and so are rather ineligible candidates for reference. Nor do our dispositions to apply moral predicates single out one particular candidate: after all, wherever there is a disagreement, there is dispo- sitional divergence. (Nor can we reasonably hope that in the long run there would be dispositional convergence-after, say, several cups of coffee and a long friendly chat. ) Where dispositions gesture at a rather eligible property, we can let the lines of that natural joint determine reference in those places

I4

I s

See. for example, ‘New work for a theory of universals,’ reprinted in Papers in Metaphysics and Epistemology, Cambridge University Press, 1999. ‘Gruesome’ is Lewis’ favorite term for highly unnatural properties. The term, though not the naturaVunnatura1 contrast, is inspired by Goodman’s famous discussion of induction in Fact, Fiction and Forecast (Harvard University Press, 1955).

SYMPOSIUM 177

where dispositions run out, or where dispositions conflict. But supposing goodness to be gruesome, what could possibly make for successful reference to that very property?16

To my mind, these concerns raise the most pressing kind of problem for moral realism. They also make vivid what a satisfying moral realism would look like, viz., some compelling package of moral and semantic theory that sketches a metaphysical landscape of the good and sketches some appropriate mechanism whereby we succeed in bringing that landscape before our minds. Perhaps a reductive naturalism is all wrong. Perhaps there are emergent mani- fest properties, including moral properties, that command high eligibility despite the gruesome nature of their supervenience base.” Or perhaps some very different theory about reference than any just sketched is the right one. Or perhaps mind-world relations are so primitive-Brentano was right-that it is wrong to seek out a theory of how successful reference occurs when it does. Or perhaps we should take seriously a moral realism that embraces semantic indeterminacy whenever there is stable disagreement between informed moral subjects, restricting moral truth to points of convergence. Or perhaps epistemicist accounts of vague concepts can help to bolster moral realism. I offer no tentative proposals of my own here. Instead I merely wish to register my puzzlement at Gibbard’s avoidance of questions such as these. I presume that he believes that it is not incumbent on his particular brand of practical realism to address them. But if not, why not?

l6 A causal theory of reference won’t help much either-for it is hardy plausible that the highly genymandered combination that the naturalist moral realism takes to constitute goodness would enjoy any special causal ‘oomph’: of the myriad combinations of natural properties, there will by many reference candidates for moral predicates, no one of which is better causally connected than the others to paradigmatics deployments of moral predicates. Cf. “The Authority of Affect” by Mark Johnston in Philosophy and Phenomenological Research 63 (2001).

178 JOHN HAWTHORNE


Recommended