DRAFT (6-18-09); Please don’t cite without permission

Metaethical Contextualism Defended
Gunnar Björnsson & Stephen Finlay

Short version (Finlay) for RoME 2009. Note that the Kolodny & MacFarlane paper is still work in progress, and is discussed here with their permission. Please consult with them before citing.

ABSTRACT:

According to metaethical contextualism, by uttering the same ‘ought’ sentences different speakers can express different propositions, because ‘ought’ is semantically incomplete and has one or more open argument-places that can be filled in different ways. There are reasons to think that ‘ought’ claims are relativized both to (i) bodies of information, and (ii) standards (or ends). Contextualism about information-relativity and contextualism about standard-relativity both face parallel objections stemming from their treatment of our practices of disagreeing across contexts. I argue first that contextualists can and should respond to Niko Kolodny & John MacFarlane’s attack on contextualism about information-relativity by adopting a certain pragmatic strategy. Then I show that structurally the same strategy is open to the contextualist as a response to Weatherson’s attack on contextualism about standard-relativity.

According to one form of relativism about normativity, different people commonly mean and say different things in uttering the same basic normative sentences, like ‘X is wrong’ and ‘One ought to do X’. Metaethical Contextualism is the semantic doctrine that terms like ‘wrong’ and ‘ought’ are semantically incomplete, and have one or more open argument-places that can be filled in different ways, so that utterances of ‘X is wrong’ and ‘X is not wrong’ do not necessarily express propositions that are contradictory.

Gunnar and I are metaethical contextualists.[1] Wethink we have a lot of good reasons for holding this view, but since my time today is short, let’s not worry about what those reasons are. What I’m interested in discussing instead is whether there is any defensible way to be a metaethical contextualist, in the face of a group of objections that have recently been pushed forcefully against contextualism by several philosophers. These objections challenge contextualism’s ability to handle our practices of normative disagreement.

I haven’t yet said what we take normative terms to be relativized to. Actually we believe in two distinct forms of relativity: to information and to ends (although for present purposes I will speak more vaguely and ecumenically about standards). On our contextualism, therefore, any meaningful utterance of a sentence, ‘A ought to ’ will express a proposition to the effect that A ought-relative-to-some-information-i-and-some-standard-s to . Parallel objections from disagreement are pressed against contextualism about both standard-relativity and information-relativity. Since I have the good fortune today to have Brian Weatherson as my commentator, I am going to focus on the challenge against standard-relativity—since Brian has recently raised that challenge in a particularly acute way. But to get there I will take a detour through information-relativity. Here, the challenge to contextualism has so far been raised only by Niko Kolodny and John MacFarlane, in work still in progress. My plan of attack is as follows. First we argue that contextualists can and should respond to Kolodny & MacFarlane’s attack on information-based contextualism by adopting a certain pragmatic strategy. Then we show that structurally the same strategy is open to the contextualist as a response to Weatherson’s attack on standard-based contextualism.

1. Information-relativity: deliberation and advice

When an agent deliberates to a conclusion about what he ought to do, it is generally the case that this ‘ought’ is sensitive to the limitations of the agent’s epistemic situation. But when advisors offer their advice as to what agents ought to do, these ‘oughts’ are generally not sensitive to the agent’s epistemic limitations, but rather are sensitive to the information of the advisor.

To use Kolodny & MacFarlane’s illustration,[2] suppose that ten miners are trapped together in one of two mine shafts, A and B, their lives threatened by impending flooding. Agent does not know whether the miners are all in A or are all in B, and is able to block either one of the shafts with sandbags, but not both. His evidence suggests that if he blocks the correct shaft then all the miners will be saved, if he blocks the incorrect shaft then all the miners will drown, and if he leaves both shafts unblocked then only one miner (who is deepest in the shaft) will drown and the other nine will survive.

The appropriate outcome of Agent’s deliberation, it seems, is the judgment that he ought to leave both shafts unblocked. A simple contextualist account can interpret this as acceptance of the proposition that Agent ought-relative-to-Agent’s-information to leave both shafts unblocked.

But now consider Advisor, who is privy to all Agent’s information but has additional information that suggests that an attempt to block shaft A will fail, flooding both shafts equally so that one miner will drown and nine survive, whereas blocking B will save all the miners if they are in B but drown them all if they are in A, while if both shafts are left unblocked then all will survive if they are in A but all will drown if they are in B.

The appropriate advice for Advisor to offer, it seems, is the judgment that Agent ought to block shaft A. A simple contextualistaccount can interpret this as acceptance of the proposition that Agent ought-relative-to-Advisor’s-information to block shaft A.

What is the alleged problem for contextualism? We can distinguish two challenges: a problem of practical integration and a problem of semantic assessment.

1.1 The Integration Problem

The problem of practical integration is that on contextualism, Advisor’s advice doesn’t answer to the deliberative question with which Agent is concerned. Agent deliberates over the question, ‘What ought I to do, relative to my information?’ while Advisor gives him an answer rather to the question, ‘What ought I to do, relative to Advisor’s information?’ Kolodny & MacFarlane observe that advice ‘is characteristically intended to help the advisee arrive at the correct answer to the question about which he is deliberating’. But on contextualism, Agent doesn’t need any help answering his question, and Advisor doesn’t provide any. Practices of deliberation and advice seem to come apart in a puzzling way.

This problem only arises for contextualism, of course, if the contextualist relativizes the ‘oughts’ of deliberation and advice to different bodies of information. As Kolodny and MacFarlane acknowledge, more sophisticated forms of contextualism are less vulnerable. Contextualists need not understand normative judgments as inflexibly relativized always to the speaker’s information, but can and should rather allow that they can be relativized to bodies of information defined in different ways, as suits the conversational purposes of the speaker. There are various ways of implementing this idea. If Agent is in the company of others, such as Advisor, then it would be sensible for him to see his practical problem as a shared problem, to be addressed through the collective resources of his audience and himself. Plausibly, Agent’s judgment is better interpreted as having the content that I ought-relative-to-OUR-information to block neither shaft. But this judgment, relativized to the information collectively shared by (at least) Advisor and himself, is false in virtue of Advisor’s information, and straightforwardly contradicted by Advisor.

Kolodny and MacFarlane do not believe that this manouvre saves the contextualist, because of the problem of ‘advice from unexpected sources’. Suppose that after Agent reaches the judgment that he ought to block neither shaft, Physicist unexpectedly lands in a helicopter, possessing knowledge that the miners are in fact all in shaft A, and she tells Agent that he ought to block shaft A. This case presents the contextualist with a dilemma: either Physicist is a member of the relevant group so that Agent’s judgment is sensitive to her information, or else she is not. Kolodny & MacFarlane believe that contextualists face problems on both horns of the dilemma.

Suppose (a) the contextualist says that Agent’s deliberation and judgment was in fact sensitive to Physicist’s information. ThenAgent’s judgment was false, and K&M suggest that he would have been ‘unwarranted and irresponsible’ in drawing it. For on a contextualist interpretation of Agent’s judgment broad enough to encompass unknown and absent people like Physicist, Agent would have no grounds for confidence that he knows what the relevant body of information contains, and to reach any conclusion would in his circumstances be unwarranted. But (as we agree) Agent’s judgment does not seem unwarranted and irresponsible.

So suppose rather that (b) the contextualist says that Agent’s judgment concerned a proposition that was not sensitive to Physicist’s information. Then we have the original problem: it seems that Physicist’s advice does not address the question to which Agent sought an answer in his deliberation.

We think that contextualists can embrace both horns of this dilemma. First, there are plausible ways in which Agent could have intended to refer to a body of information that included what Physicist knew. Consider as an illustration what we will call news-sensitive contextualism. Faced with a practical problem, agents usually have a window of time to conduct research and gather information before the moment when they have to act. It is implausible, therefore, that the question a deliberating agent asks would be ‘What ought I to do, relative to my present information?’ More plausibly, deliberating agents are concerned with what they ought to do, relative to something like the best body of information they are able to acquire by the time when they must decide what to do.[3] Since Physicist makes her information available in time, Agent’s judgment would then be sensitive to it. While Agent would then have judged falsely, we disagree with Kolodny & MacFarlane that he would have judged irresponsibly. It was perfectly reasonable for Agent to have expected (mistakenly) that no information like Physicist’s would be forthcoming in the short time at his disposal.

However, this solution obliges us to find a solution to the other horn of the dilemma too. For on this contextualist account, the sensitivity of Agent’s judgment to Physicist’s information depends on whether Physicist chooses to make it available. Hence, Physicist makes Agent’s judgment false by making her information available. We still need to explain how Physicist is helping Agent answer the question he deliberates over.

It turns out, however, that this is everybody’s problem. It’s not at all plausible to respond to this kind of case by denying that the truth of an ought-judgment like Agent’s is information-relative. (As Agent well knows, what he ought to do given all the facts both known and unknown is to block the shaft that actually contains the miners. But knowing that shouldn’t stop him from reaching and acting on the deliberative judgment that he ought to block neither shaft.) Kolodny & MacFarlane argue that it should lead us to embrace the radical thesis of semantic relativismin place of(orthodox)semantic contextualism: the thesis that although different ought-claims like those made by Agent and Physicist express nonrelativized propositions (contra contextualism), these propositions have different truth values for different assessors. The proposition that Agent ought to leave both shafts unblocked is true as assessed relative to Agent’s perspective and information, and false as assessed relative to Physicist’s perspective and information.

But the move from contextualism to relativism does not help to resolve the integration problem. Although Agent and Physicist are both interested in the truth values of the same proposition, they are not now interested in the same truth value. Agent deliberates with purpose of determining the truth relative to Agent’s information of the proposition that he ought to leave both shafts unblocked, while Physicist advises him rather about the truth relative to Physicist’s information of that proposition. Any perceived advantage here to relativism over contextualism is therefore illusory.

We suggest that the solution to the integration problem lies in rejecting a basic assumption underlying Kolodny & MacFarlane’s challenge: the assumption that the fundamental aim of deliberation is to determine the truth of certain ought-propositions. This assumption doesn’t take account of the fact that deliberating agents generally prefer better and fuller information if they can get it. Rather, we suggest that agents’ fundamental aims in deliberating are typically the promotion and protection of certain values or things that matter to them.[4] In the mine scenario, Agent’s fundamental concerns are for the preservation of each of the miners’ lives. Fuller information is desirable to him because it increases the likelihood that he can make better decisions with respect to promoting those values.

Given this view of the aims of deliberation, determining the truth of ought-propositions relativized to particular bodies of information will be of derivative and instrumental interest to deliberating agents. Deliberators seek to determine what they ought to do relative to the best information available to them, simply because given their epistemic situations, these propositions are the best guides available with respect to their preferences. But should they become privy to new information, the former proposition loses this derivative significance and becomes pragmatically moot. Deliberating agents would therefore lose all interest in them, and abandon them for the question of what they ought to do relative to the new, improved body of information.

Hence, Agent is fundamentally concerned to prevent the deaths of each of the miners, and only has an instrumental interest in determining what he ought to do relative to his own information. But he would be very much interested in acting instead on an ought-judgment based on superior information. This is something that Physicist is in a position to provide, and hence Physicist’s advice assists Agent address his fundamental practical concerns.

1.2The semantic assessment problem

Even if contextualists can explain how judgments of deliberation and advice could be integrated, it doesn’t follow that this is how deliberation and advice are actually integrated. Contextualism’s opponents maintain that our linguistic practices are inconsistent with what contextualism predicts—particularly our practices of semantic assessment, or responding to ought-claims with ‘Yes’ or ‘No’, and our evaluations of them as true or false.

Here we’ll have to make do with a single example. When Physicist unexpectedly arrives, she might plausibly reject Agent’s claim that he ought to leave both shafts unblocked by saying,

(1)No, you ought to block shaft A.

This looks like a problem for contextualism. If Agent’s judgment was relativized to his own information, then it was true, and Physicist has no grounds for thinking otherwise. If Agent’s and Physicist’s ought-claims are relativized to different bodies of information, then they are not contradictory. But intuitively it seems that unexpected advisors with better information often do disagree with agents’ ought-claims. For example, it would be very strange for Physicist to say,

(2)Yes, that’s true. But you ought to block shaft A.

This supports the semantic relativist’s claim that the content of ought-claims is not relativized to bodies of information.

News-sensitive contextualism can accommodate this point about advice, since it identifies a common subject-matter. But a similar problem reemerges for any form of contextualism from cases of mere evaluation or eavesdropping, in which the assessor is unable to provide her information or advice. Suppose that Physicist is observing Agent over closed-circuit television, with no way to communicate with him. Hearing him declare that he ought to leave both shafts unblocked, she might say,

(3)No, he ought to block shaft A.

Again, it would be very strange for her to say,

(4)Yes, that’s true. But he ought to block shaft A.

On its face, it therefore appears that semantic relativism treats our agreement and disagreement with ought-claims more faithfully than even a sophisticated version of contextualism does.

This objection against contextualism relies on a covert but significant assumption: that appropriate assessment of a person’s utterance or judgment with ‘yes’ or ‘no’, and as ‘true’ or ‘false’, is always assessment of the same proposition that the person asserted or accepted. As an empirical matter, this assumption is mistaken. Consider,

(5)[A:] I think that Sally stole the money.

(6)[B:] ??Yes, that’s true, you do.

(7)[B:] Yes, that’s true, she did.

(8)[A:] I can’t believe how healthy John looks.

(9)[B:] ??Yes, that’s true, you can’t.

(10)[B:] Yes, I can’t either.

It is natural to respond to an utterance of (5) or (8) by offering an assessment of a proposition other than that which (on an orthodox account) they literally assert. Often, as with (5), the assessed proposition is also plausibly something that the speaker implicated or in some other way expressed. But in other cases the assessed proposition is apparently not plausibly anything that the speaker expressed. In the case of the assessment of (8), it seems most plausible that the ‘Yes’ of (10) is meant as acceptance of the proposition that B can’t believe how healthy John looks.

This observation may seem surprising, and calls for explanation. How could it be that the salient proposition for the semantic assessment of people’s speech or mental acts is a proposition other than that which they thereby asserted or accepted? We suggest that propositions are selected for salience pragmatically, in accordance with the following principle:

Pragmatic Principle of Assessment: The salient proposition for semantic assessment of a speech or mental act is the proposition associated with that act that has the most relevance for conversational interests.

In the case of (5), what is of primary conversational interest is the truth of the proposition that the speaker reports that he believes, and not the reported fact of his believing it. In the case of (8), we suggest that by sharing his opinion in the mode of gossip, the first speaker implicitlyinvites the other to reciprocate by sharing his own opinion on the same subject.[5] Of course, usually the most conversationally relevant proposition associated with a speech act will be the original proposition asserted. But we see that this is not invariably the case. In section 1.1 above, we found reason to think that it is often not the case with ought-claims in particular.