«© Φ Philosophical Writings Proceedings of the Fifteenth Annual British Postgraduate Philosophy Conference EPISTEMIC CONTEXTUALISM, EPISTEMIC ...»
© Φ Philosophical Writings
Proceedings of the Fifteenth Annual British Postgraduate Philosophy Conference
EPISTEMIC RELATIVISM AND
Robin McKenna – University of Edinburgh
In the recent philosophy of language literature there is a debate over whether contextualist accounts of the semantics of various terms can accommodate intuitions of disagreement in certain cases involving those terms. Relativists such as John MacFarlane have claimed that this motivates adopting a form of relativist semantics for these terms because the relativist can account for the same data as contextualists but doesn’t face this problem of disagreement (MacFarlane 2005, 2007 and 2009). In this paper I focus on the case of epistemic predicates and I argue that on a certain assumption about what is involved in assessing an utterance the epistemic contextualist can solve her problem of disagreement. This undercuts a motivation for epistemic relativism.
Introduction One night Bob, an anti-sceptic, has a dream in which he loses his hand. He wakes up, looks at his hand, and exclaims, “I know that I have a hand, that was just a dream.” Consider two sorts of cases. First, say that Bob woke his sceptical wife, Saskia, who challenges him, saying, “No, you don’t know that. You can’t rule out the possibility that you’re a handless brain in a vat.” Second, say the next day Bob explains his dream to his colleague, Enrico, who discusses Bob’s dream with the office sceptic, Marie, who says, “No, Bob doesn’t know that he has a hand. He can’t rule out the possibility that he’s a handless brain in a vat.” What I will call the basic motivation for epistemic contextualism is that
the standards so her knowledge denial is true. In Marie’s context the standards are high so her knowledge denial is true.
Contextualist semantics have been proposed for a number of expressions such as taste predicates, moral predicates, aesthetic predicates, epistemic modals and epistemic predicates. However, in the recent literature rival relativist semantics for each of these sorts of expressions have been proposed, and a main motivation for adopting relativist semantics is supposed to be that relativists can account for the same data as contextualists but don’t face the problem of disagreement.2 If Julian says “I’m tired” and Anya responds, “No, I’m not tired”, Anya isn’t disagreeing with Julian. What Julian has said is that Julian is tired and what Anya has said is that Anya is not tired. The only way of explaining why Anya has used the disagreement marker “No” is by taking Anya to be confused about the meaning of indexicals. Compare this with our cases above. Many would take it that both Saskia and Marie are disagreeing with Bob.3 But, if the epistemic contextualist is right, what Bob has said is something like Bob knows by low standards that he has hands and what both Saskia and Marie have said is something like Bob doesn’t know by high standards that he has hands. These propositions don’t contradict each other so, as in the Julian and Anya case, there’s no disagreement. Is taking Saskia or Marie to be confused about the meaning of ‘knows’ the only way of explaining why either of them use the disagreement marker “No” to negatively assess Bob’s claim? The main thesis of this paper is that the epistemic contextualist can provide an alternative explanation of these patterns of use. I take it that the challenge posed by the problem of disagreement to the epistemic contextualist is providing such an explanation. If that’s right, this undercuts a main motivation for adopting a form of relativist semantics for knowledge ascriptions.
I proceed as follows. In §1 I distinguish indexical contextualist and relativist accounts of the semantics of ‘knows’. In §2 I present Keith DeRose’s proposed solution to the problem of disagreement and an objection to that solution. In §3 I argue that the epistemic contextualist can solve the problem of disagreement.
For predicates of taste see Glanzberg 2007, Huvenes forthcoming, Kölbel 2009 and MacFarlane 2007. For moral predicates see Brogaard 2008 and Björnsson & Finlay 2010.
For aesthetic predicates see Baker forthcoming. For epistemic modals see von Fintel & Gillies 2009 and MacFarlane 2009. For epistemic predicates see Cohen 1999, DeRose 1995, Lewis 1996 and MacFarlane 2005.
In common with most of the literature, I’m going to take it for granted that we have the intuition that both Saskia and Marie disagree with Bob.
© Φ Philosophical Writings 92 Epistemic Contextualism, Epistemic Relativism and Disagreement
1. Indexical Epistemic Contextualism and Epistemic Relativism Take a standard semantic framework in which a sentence uttered in a context has a content that is evaluated for truth relative to the circumstance of evaluation of the context in which it was uttered.4 Sam’s utterance of the sentence ‘I am sitting’ at t is true if and only if Sam is sitting at t in the actual world. The time at which Sam uttered the sentence is a determinant of the truth-value of Sam’s utterance. On this standard framework, the determinants of the truth-value of an utterance either determine the content or the circumstance of evaluation relative to which the content is evaluated. Current orthodoxy is that the time at which one utters a sentence determines the content rather than the circumstance.5 So, on this view, the content of Sam’s utterance is that Sam is sitting at t and that content is evaluated for truth relative to the actual world. Circumstances of evaluation are worlds, and sentences uttered in contexts express contents that are evaluated for truth or falsity relative to such circumstances.
Say that in context c Payal utters the sentence ‘S knows that p’. For the epistemic contextualist, one of the determinants of the truth-value of Payal’s utterance is the epistemic standard operative in c. This determinant, as with the time of Sam’s utterance of ‘I am sitting’, can either determine the content of Payal’s utterance or the circumstance relative to which it is evaluated. Indexical Epistemic Contextualism (IEC) is the view that the standard determines the content.
Epistemic Relativism (ER) rejects this semantic framework. ER is the view that knowledge ascriptions have a content that is evaluated for truth or falsity relative to the world in which they are uttered and the epistemic standard operative in the context in which they are assessed.
Take Payal’s knowledge ascription ‘S knows that p’. For ER, this ascription has as its content that S knows that p and that content is evaluated for truth relative to Payal’s epistemic standard (MacFarlane 2005).
It’s something of a commonplace in the literature that ER doesn’t face the problem of disagreement.6 On ER, when Bob says that he knows he has a hand that claim is true as assessed by Bob because in his context This framework is from Kaplan 1989.
See, for example, King 2003.
Some have disputed this. See Dreier 2009 and Francén 2010.
© Φ Philosophical Writings Robin McKenna 93 the epistemic standards are low. However, that claim is false as assessed by both Saskia and Marie because, in their respective contexts, the epistemic standards are high. That explains why Saskia and Marie negatively assess Bob’s claim. Bob, Saskia and Marie are all assessing the same claim and, as assessed from Bob’s context, that claim is true but, as assessed from Saskia and Marie’s contexts, that claim is false. In this paper I’m going to assume that ER doesn’t face the problem of disagreement.
In the next section I will discuss DeRose’s proposed solution to the problem of disagreement. I will argue that his solution fails and then present what I take to be a better solution.
2. DeRose’s Single-Scoreboard Semantics
IEC is the view that the truth-conditions of knowledge ascriptions are in part determined by the epistemic standard operative in the context in which those ascriptions are made. Sentences involving the word ‘knows’ are semantically incomplete and only express complete propositions when supplemented by a contextually salient epistemic standard. On DeRose’s view, in any given context the conversational participants are meant to converge on a single epistemic standard and once that standard has been converged upon it’s that standard that allows those sentences to express complete propositions (DeRose 2009, pp. 135But, in conversations where A and B have different standards, an ascription or denial of ‘knowledge’ to S is true (or false) if and only if S meets (or fails to meet) the standards of both A and B, and truthvalueless if and only if S meets (or fails to meet) one set of standards but not the other (ibid, pp. 144-5). The idea is that when the speakers in a context can’t agree on a single epistemic standard, sentences involving the word ‘knows’ can’t express complete propositions and so remain truth-valueless.
In the Bob and Saskia case Bob has low epistemic standards whereas Saskia has high standards. So, on DeRose’s view, when Saskia says that Bob doesn’t know he has a hand that’s truth-valueless. DeRose thinks that this deals with the problem of disagreement because, to use his phrase, Saskia is disagreeing with Bob over the truth-value of the same ‘gappy’ thing (ibid, p. 145). What about the Bob and Marie case?
In such cases DeRose also holds that an ascription or denial of ‘knowledge’ to S is true/false if and only if S meets/fails to meet the © Φ Philosophical Writings 94 Epistemic Contextualism, Epistemic Relativism and Disagreement standards of both A and B, and truth-valueless if and only if S meets/fails to meet one set of standards but not the other (ibid, pp. 148-50). On DeRose’s view, when Marie says that Bob doesn’t know he has a hand that’s truth-valueless. Again, this is supposed to deal with the problem of disagreement. Marie is disagreeing with Bob over the truth-value of the same ‘gappy’ thing.
I’m going to grant DeRose that this solves the problem disagreement in both cases. I will argue that DeRose’s solution to the problem of disagreement is incompatible with the basic motivation for epistemic contextualism. Recall how the Bob and Marie disagreement went. Bob has just woken from a dream in which he lost a hand. Upon awakening he affirmed, “I know that I have a hand, that was just a dream.” Later, Marie is informed of Bob’s remark and says, “No, Bob doesn’t know that he has a hand.” Consider what would happen later if Bob, still in a low standards context, recalls his dream and says, “I know that I have a hand, thank God.” Presumably, on DeRose’s view this assertion is truth-valueless because it meets Bob’s standards but not Marie’s. DeRose’s view has the consequence that, for any knowledge ascription, if that knowledge ascription is disputed then there is no context in which that knowledge can be truly re-ascribed. But the basic motivation for epistemic contextualism is that in low standards contexts, like Bob’s when he self-ascribes the knowledge that he has hands, a good number of knowledge ascriptions are true, whereas in high standards contexts, like Marie’s when she challenges Bob’s knowledge selfascription, a good number or maybe all knowledge denials are true.
Because on DeRose’s view disagreement renders knowledge ascriptions or denials truth-valueless irrespective of the context in which those ascriptions or denials are made, it is incompatible with the basic motivation for epistemic contextualism.
3. An EIC Solution to the Problem of Disagreement In this section I’m going to argue that, on an assumption about what one is assessing when one assesses an utterance, EIC can solve the problem of disagreement.
In their paper ‘Contextualism, Assessor Relativism, and Insensitive Assessments’, Gunnar Björnsson and Alexander Almér identify the
following implicit assumption about utterance assessment:
When we assess utterances using various assessment phrases, we normally (barring confusion, misunderstanding, etc.) assess the satisfaction of their truth-conditions. (Björnsson and Almér 2009, p. 366) So, on this assumption, when Saskia or Marie assesses Bob’s utterance, they are assessing the satisfaction of the truth-conditions of Bob’s
utterance. Björnsson and Almér propose a revision of this assumption:
Note that, of course, it may well be that the conditions made most salient by an utterance are just the truth-conditions of the utterance, as in
the example below:
Here Tony is assessing the truth-conditions of Kyle’s utterance.
Presumably, on Björnsson and Almér’s view, this is because that’s what is made most salient by the utterance in Tony’s context of assessment.
I’ll quickly run through their argument for revising the assumption and their argument for their proposed revision.7 They identify certain cases where it looks like the assessment of an utterance is not assessing
the satisfaction of the truth-conditions of the utterance. For example:
If we accept the plausible claim that one’s claims about what ‘might’ be the case are relative to one’s present body of information, or possibly the body of information of some relevant group, Bill’s utterance is true if and only if, relative to the information he possesses, it’s possible that the keys are in the car.8 So Emily isn’t assessing the truth-conditions of Bill’s utterance. Rather, as Björnsson and Almér have it, she’s assessing