It has recently been said that epistemology is experiencing a “value

advertisement

Forthcoming in Greco, J. and D. Henderson (eds.) Epistemic Evaluation (OUP)

Two Purposes of Knowledge Attribution and the Contextualism Debate

Matthew McGrath

University of Missouri

In this paper, I take up Edward Craig’s (1990) advice to ask what the concept of knowledge does for us and treat our answers as clues about its application conditions. What a concept does for us is a matter of what we can do with it, and what we mainly do with concepts is deploy them in thought and language. So, one natural place to look to find out what the concept of knowledge does for us is at the purposes we have in attributing knowledge. This paper examines two such purposes, agent-evaluation and informant-suggestion, and brings the results to bear on one important debate about the application conditions of the concept of knowledge – the debate between contextualism and its rivals.

§1. First purpose: agent evaluation

It is a familiar feature of daily life that we criticize and defend actions by attributing knowledge. Examples are easy to come by. The tub has a drip. Your spouse is not particularly handy but refuses to call the plumber. He’s applying all the force he can muster to the cold water handle. He keeps trying and trying. You finally say: “you know that’s not going to work! Let’s just call that plumber, dear.” Here you criticize your spouse’s action by attributing a piece of knowledge. Or, to use an example of Stanley’s, suppose you and your spouse are driving to a restaurant in a city you’re visiting. At an intersection, you turn left toward a residential

neighborhood instead of right toward a row of restaurants. Your spouse looks at you worriedly.

You say, “It’s ok. I know the restaurant is this way.”

Nor do we cease to attribute knowledge for these evaluative purposes when the stakes are high. On Friday, when you see that the lines are long at Keith DeRose’s bank, and you really must have that check deposited by Monday, then you can defend your not waiting in the long lines to deposit the check today (Friday) by saying, “it’s okay; I know it’s open tomorrow.” It might be harder to get away with a knowledge-claim in high-stakes practical environments, but they can still be used to defend one’s action. And to criticize as well. If Keith keeps checking and checking, at some point we will criticize him, “ok, Keith; now you know!”

These examples given so far are first- and second-person present tense. But we can and do evaluate actions by using third-person and past knowledge-attributions. On Monday, after the check as cleared, one might defend Keith’s waiting to deposit it on Saturday by saying, “he was fine to wait; he knew it was open the next day.” Keith might of course do so himself as well. 1

My son might criticize me under his breath for continuing to apply force to the cold water handle by saying, “Daddy knows this isn’t going to work” and he certainly can criticize me after it breaks – by saying, “Daddy knew it wouldn’t work! He shouldn’t have kept trying!”

In each of these examples, knowledge is being attributed to an agent as a basis for an evaluative judgment about the agent’s

-ing. The judgment may be a simple up or down on

ing, as it is the examples I gave above – someone “should” or “shouldn’t”

or have

-ed” – but it also might be more nuanced – there’s a “reasonably good case” for the person to

or not to

.

1 All these examples involve attributions rather than denials of knowledge. It’s true that sometimes we do defend and criticize action by denying knowledge, and that we often defend and criticize belief this way. If we want to go to a certain neighborhood X in a city, we might say, “We can’t just get on this bus; we don’t know it goes to X.”

And certainly we sometimes say, in response to others’ professed beliefs, “you don’t know that.” The matter merits attention, but I will consider only attributions of knowledge here.

2

So, I might say, “well, she knows the train costs twice what the bus does, so that is a factor in favor of the bus.” I will focus on up or down judgments.

Knowledge-attributions can thus be used as bases for assessments of agents’ actions. But is this a feature which sets the concept of knowledge apart from other concepts? As Jessica

Brown (2008b, 1139-40) remarks, one can mention just about anything on some occasion or other to accomplish such evaluative purposes. To use one of her examples, I can mention a friend’s plan to run a marathon to criticize the cook’s not giving her a larger portion of the soup or to defend my giving her a larger portion.

It’s true that when a first- or second-person present-tense knowledge-attribution is used for agent-evaluation one could often accomplish the same purpose by simply asserting the relevant proposition – e.g., “she is running a marathon later today; you should give her a bigger share of the soup!” (Arguably, this has much to do with the relations between assertion and knowledge.) But when one is speaking about an agent not present in the conversation or speaking about an action in the past, very often the simple “P” will not be effective for evaluative purposes, because much will depend on the agent’s epistemic or psychological state regarding P.

If we arrange that you’ll pick me up from the airport at 8pm, then if I arrive early at 6pm I can hardly mount a serious criticism of you for not being there when I get off the plane. I can say, “I was there at 6pm and had to wait”, meaning this as a criticism, but it is all too easy for you to undermine it by appealing to your lack of awareness: “I had no idea; you didn’t tell me.”

Compare the criticism, “you knew I was arriving early!” Agent-evaluation in such cases won’t stick, so to speak, unless it is based on an attribution of some epistemic or psychological relation between the agent and P.

3

Still, this at best places the concept of knowledge in a large class of concepts – having some reason/excellent reason to believe P, believing P, being sure that P, etc. It doesn’t set it apart. But we can set it apart, as I’ll next argue.

Consider a bank case, which could be high- or low-stakes. Keith decides to wait to deposit the check tomorrow, rather than standing in the long lines to deposit it today. Suppose, in defense of Keith’s waiting, I point to Keith’s having “good reason to think” it’s open tomorrow. I have in effect given a little argument for my positive assessment of Keith’s action:

“Keith has good reason to think the bank is open tomorrow, so he can just wait to deposit it tomorrow.”

Suppose you reject the conclusion of this little argument, and suppose in particular you reject it because you think some evaluative/normative contrary of it is true. So, suppose you claim Keith can’t wait till tomorrow; he must stand in line today.

2

How might you object to my argument?

You could reject its premise, of course. Alternatively, you could claim that, although the premise is true, still, the conclusion is false. That is, you could give a response of the form:

“Yes, Keith has good reason to think the bank is open tomorrow, but no, he can’t just wait till tomorrow to deposit the check.”

I want to investigate this sort of response further. On what basis might one make this response?

I want to discuss two.

One is on the grounds of what I’ll call deliberative weight. So, for instance, you might agree that Keith has good reason to think the bank is open tomorrow but insist he can’t just wait till then because, for instance, his relatives will be visiting tomorrow and it would be too much of an inconvenience to try to get away to make the deposit, or perhaps because the lines are often even longer then. You might also mention epistemic considerations concerning the relation

2 Here ‘can’t’ expresses impermissibility rather than inability. As Chase Wrenn points out to me, grammarians tend to discourage this usage.

4

between the bank’s being open tomorrow and Keith’s depositing it then, e.g., by saying “Keith doesn’t know he could deposit it tomorrow even if the bank is open then.” In offering any of these responses, you are in effect suggesting that the issue of whether the bank is open tomorrow isn’t or shouldn’t be the decisive issue for Keith in his decision-making. In effect, you’re suggesting that even if you “give” Keith the bank will be open tomorrow as a premise to work with; this wouldn’t settle the question for him of what to do.

3

Another basis on which you might accept the premise but deny the conclusion concerns the agent’s epistemic relations to the proposition in question. You might claim that although

Keith has good reason to think the bank will be open tomorrow he shouldn’t wait till tomorrow because he needs better reason/evidence/grounds to think it’s open tomorrow if he is to wait to deposit the check then rather than standing in line today. You might say, “Yes, Keith has good reason to think it’s open tomorrow; but he shouldn’t wait till then because he doesn’t have good enough reason to think it’s open tomorrow.” Let’s call this way of responding to my little argument the epistemic objection .

I do not mean to imply that the epistemic objection can only be raised using what might be considered internalistic epistemic concepts (how strong one’s reasons/evidence/grounds are).

Epistemologists often use a generalized notion of “epistemic position” to cover all these epistemic factors, including those that are externalistic, e.g., how reliable one’s p-relevant beliefproducing processes are, how safe one’s indications of p are. One way to raise the epistemic objection against the above defense of Keith’s waiting to deposit the check Saturday is to concede that Keith has good reason to think the bank will be open then but to claim that he shouldn’t wait till then because he doesn’t have strong enough epistemic position with respect to

3 Whether a particular “deliberative weight” objection succeeds will depend on the sort of evaluation at issue, whether it is an evaluation of the agent’s reasonableness or of something more objective.

5

the bank is open tomorrow

. One might do this, for instance, by saying, “yes, Keith has good reason to think the bank is open, but he shouldn’t wait till tomorrow, because the bank is closed tomorrow!

”, 4

or “because his good reason is based on a report from a friend who was mixing up

Keith’s bank with another one.”

In general, where ‘V’ is a cognitive or epistemic verb phrase, and ‘ 

-ing’ picks out an action or omission, let’s say that to raise the epistemic objection to the argument:

S Vs that P, so S can

 is to accept that S Vs that P but claim that S can’t – or ought not/mustn’t/shouldn’t –

because

S doesn’t have strong enough epistemic position with respect to P. Epistemic objections to other

V-based evaluations such as the below are understood accordingly:

S Vs that P, so S must

.

S Vs that P, so S mustn’t

.

S Vs that P, so S need not

. as well as past tense versions of these.

Clearly, the epistemic objection is often made in response to evaluations based on attributions of many cognitive, epistemic and alethic states . But now, suppose I defend Keith’s action by saying,

“Keith knows the bank is open tomorrow, and so he can wait to deposit the check tomorrow.”

One doesn’t in ordinary life go on to dispute arguments like this by raising the epistemic objection. One doesn’t, first, concede that Keith knows that the bank is open tomorrow, and then insist with no air of retraction that still he needs better grounds before waiting to return tomorrow. Indeed, a speaker might use ‘knows’, and especially the stressed ‘ knows’

, precisely to

4 Thanks to Adrian Haddock.

6

rule out this sort of objection. You can stress ‘ good reason’ or ‘ excellent reason’ all you like and still the epistemic objection can be and often is made.

Return to the tub and its drip. We can sense, I think, that the criticism “You know that applying force won’t work! You should call the plumber” is stronger in some way than the criticism “You have good reason to think that applying force won’t work! You should call the plumber.” We now can see, in a rough and ready way, why this would be: whereas the latter leaves some wiggle room epistemically, the former does not. It’s open to me to reply, “I do have good reason to think applying force won’t work, but I don’t need to call the plumber; it might well work, and it is worth a shot trying.” Here I in effect concede I had good reason but deny I had decisive enough grounds to make continuing to apply force unreasonable. But with the knowledge-based criticism, I cannot get off the epistemic hook in this way. I can’t say, “Yes, I know it won’t work, but it might well work, and so it is worth a shot.” Similarly, suppose in another “tub case” I defend my choice to replace the washer rather than call the plumber by saying “I have good reason to think the washer is problem; I don’t need to call the plumber.” It is open to my spouse to reply, “Well, perhaps you do have good reason, but not good enough; it might very well not work, and then what a waste of time and effort!” By defending my action using “know” I am not allowing myself to be put on the epistemic hook in this way. My spouse can’t say, “yes, you know the washer is the problem, it might very well not work, and then what a waste of time and effort!”

Knowledge-attributing agent evaluations thus appear to “close off” the epistemic objection. In saying this, I don’t mean merely that people don’t as a matter of fact respond to a knowledge-attributing defense or criticism with the epistemic objection. They don’t, I think, and that is evidence of closing off, but not what it is to close off the objection. I will not try to give

7

an account of this notion of “closing off.” I hope it is enough to note that in the case of knowledge-attributing agent-evaluations, it is not only misplaced but somehow confused to make the epistemic objection.

I expect some philosophers will say that they find nothing misplaced or confused in making the epistemic objection in certain high-stakes cases. They might point to examples from

Jessica Brown (2008a), such as:

SURGEON

A student is spending the day shadowing a surgeon. In the morning he observes her in clinic examining patient A who has a diseased left kidney. The decision is taken to remove it that afternoon. Later, the student observes the surgeon in theatre where patient

A is lying anaesthetised on the operating table. The operation hasn't started as the surgeon is consulting the patient's notes. The student is puzzled and asks one of the nurses what's going on:

Student: I don't understand. Why is she looking at the patient's records? She was in clinic with the patient this morning. Doesn't she even know which kidney it is?

Nurse: Of course, she knows which kidney it is. But, imagine what it would be like if she removed the wrong kidney. She shouldn't operate before checking the patient's records.

(2008a, 176)

To bring this example squarely into our paradigm, let’s modify the dialogue as follows, so that the nurse is clearly raising the epistemic objection in response to a knowledge-attributing criticism:

Student: The surgeon knows it’s the left kidney, so she can operate without delay. She doesn’t need check the charts.

Nurse: Yes, she knows it’s the left kidney, but she can’t operate without delay. She needs to check the charts. Imagine what it would be like if she removed the wrong kidney. She would need stronger evidence to operate straightaway.

This is an important objection. I will offer two responses to it.

First, notice that there is a marked difference between the epistemic objection in response to this knowledge-attributing criticism and the epistemic objection made in response to criticisms

8

based on attributions of any number of other psychological and epistemic concepts, even ones that ascribe a high degree of evidence or epistemic position – higher than many moderate invariantists such as Brown think is required for knowledge. Compare the above speeches with variants involving attributions of having good/excellent reason. With these, the epistemic objection can be perfectly in order; with knowledge, something seems clearly worse about it.

And this contrast is in place even for certain factive notions (some of which might be thought contenders for an account of knowledge under moderate invariantism). Compare these two epistemic objections to a past-tense knowledge-based evaluation:

“Yes, the surgeon had a reasonable true belief (or: highly reasonable true belief) that it was the left kidney, but she shouldn’t have operated straightaway because she needed stronger evidence for that belief.”

“Yes, the surgeon knew that it was the left kidney, but she shouldn’t have operated straightaway because she needed better evidence for her belief.”

The latter is much worse than the former.

Second, I would like to make two empirical claims, ones which I, as a long-standing and highly competent speaker of English (!), feel quite entitled to make. The first is: we don’t regularly raise the epistemic objection in the case of ‘knows’ even in high stakes cases like this

SURGEON case. We do sometimes fail to engage the question of knowledge by asking “but can she be sure?” And we sometimes refuse to engage in evaluation and insist on explanation:

“maybe she does know it, but she isn’t going to operate because she feels unsure.” And of course we raise concerns about protocol or rules. What we don’t do is to agree firmly and in an ordinary flat “factual” tone of voice that the agent knows and then, without any air of retraction, go on to object to the assessment of the agent’s action on grounds that her evidence or epistemic position isn’t strong enough. We do all this regularly for a wide variety of psychological and epistemic notions. If the epistemic objection wasn’t closed off, we would expect that epistemic

9

objection would be widely made in high-stakes cases like SURGEON. The fact that it isn’t is evidence that it is closed off.

The other empirical claim is this: we regularly defend our actions in high-stakes cases against charges that they are “too risky” by claiming to know.

Thus, if someone questions

Keith’s waiting to deposit the check Saturday, Keith might well say

– and people in his situation do often say – “Look, I know it’s open Saturday!” But by doing this, Keith would seem to be just asking for the epistemic objection if it could be made, because here clearly very strong evidence or grounds is required. Or if “just asking” for the objection is too strong, Keith would at least be inviting questions about how strong his evidence or grounds are. But this is precisely what he isn’t doing. He is, rather, attempting to settle the question of whether his evidence or grounds is enough.

5

So, my verdict on SURGEON is that it fails to undermine our claims about what is special about knowledge-attributions as bases for assessments of actions.

I have argued that knowledge-attributing agent evaluations close off the epistemic objection. And I have argued that in this respect knowledge-attributions differ from attributions of many other psychological and epistemic notions. I now want to propose an explanation of this, while yet remaining neutral about the truth-conditions for knowledge-attributions. Here is what I claim is going on in the case of ‘knows’ but not for the other psychological or epistemic notions we’ve discussed. By attributing knowledge that P as a basis for the evaluative judgment concerning

-ing, the speaker is treating the agent as warranted enough in P for P to be a reason the agent has to

(or not to

); so anyone who accepts the speaker’s knowledge-attribution

5 A third response, which I do not have room to develop fully here, is that raising the epistemic objection can, in certain not unusual cases, commit one to making further very problematic knowledge attributions, e.g., “the surgeon knows both that there is more reason to operate straightaway but that she must not operate straightaway.”

10

thereby commits herself to treating the agent likewise.

6

This “treating” need not and usually will not take the form of an explicit verbal acceptance of a proposition to the effect that the agent is warranted enough to have P as a reason. Rather, to treat an agent this way is to be prepared to think and reason in ways that assume or presuppose that the agent doesn’t need any stronger warrant for P for P to be among her reasons. And this means, among other things, not raising the epistemic objection.

7

Thus, to raise the epistemic objection in the case of knowledge-attributing evaluations is to take on a commitment to treat an agent a certain way and then straightaway violate that very commitment. This, I suggest, is what makes the epistemic objection misplaced and confused for knowledge-attributions.

In summary, I claim the concept of knowledge serves the following “agent-evaluative” purpose: we often want to give grounds for the assessment of an action in terms of a subject’s epistemic relation to P; and we would like to be able, in many cases, to close off the epistemic objection; attributions of the concept of knowledge allow us to do this. The concept of knowledge serves this purpose because in attributing knowledge that P as a basis for assessment of an agent’s action one’s knowledge-attribution functions to commit anyone who accepts one’s knowledge-attribution to treating the agent as warranted enough in P for P to be among her reasons for or against taking that action.

I don’t claim that only knowledge-attributions can serve the agent evaluative purpose.

Consider: having conclusive evidence/grounds for P, being justified in taking P as settled. I

6 ‘Warranted enough’ is essential here. Even if your attribution of knowledge that P in defending an agent’s

-ing is accepted by a conversant, the conversant need not agree that P was a genuine reason the agent had to

(or not to

). Think of saying that you know that 2+2=4 in defense of your taking out the trash. To be warranted enough in P for P to be a reason an agent has to

is for one’s strength of warrant – or one’s strength of epistemic position – not to stand in the way of P being a reason the agent has to

. P might still not be a reason the agent has to

because P isn’t deliberatively relevant for the agent to the question of whether to 

. See Fantl and McGrath (2009, ch. 3).

7 I am relying on an assumption here, viz. that one who treats an agent as warranted enough in P for P to be among her reasons to

also thereby the agent as warranted enough in P for P to be a reason which justifies the agent in

ing, and so which makes it the case that the agent can permissibly

. This assumption is called Safe Reasons in

Fantl and McGrath (2009), ch. 3.

11

would argue that the latter should be understood in terms of having justification strong enough for knowledge.

8 As far as the former is concerned, if ‘conclusive evidence’ is understood so as to be consistent with not having evidence that justifies maximal certainty, then again it is not obvious that conclusive evidence shouldn’t be explained in terms of knowledge. So, the concept of knowledge might well figure essentially in explanations of these notions. We can say, at the very least, that in serving this agent-evaluative purpose the concept of knowledge enjoys membership in a rather exclusive class of concepts.

9

§2. Second purpose: suggesting informants.

E.J. Craig has proposed that the point of the concept of knowledge is to flag good informants. The plan of his book is to “creep up” on the concept of knowledge by examining a subjectivized version of the concept of a good informant – the concept of a good informant for me here now – and then considering the results of the objectivization of such a concept.

However, not much attention in the book is paid to the examination of the actual use of knowledge-attributions as a basis for suggesting or recommending informants. My focus here will be on this use. I will be concerned particularly with the way we use knowledge-attributions to suggest informants to particular recipients or groups of recipients.

To suggest someone as an informant to a recipient is to suggest that the recipient treat that person as an informant. There are a number of ways to do this. I might treat you as an informant for me on whether P even if I know that you are undecided on the matter yourself but have evidence bearing on the question that I don’t. (You saw next week’s weather forecast and I

8 See Fantl and McGrath (2009), chapter 5.

9 See the appendix for a discussion of whether knowledge-attributing defenses of cognitive states also close off the epistemic objection. The discussion in the appendix indirectly gives grounds for thinking that the epistemic objection is closed off because the acceptance of knowledge-attributions, when used for the two purposes, brings with it the associated reason-related commitments.

12

didn’t.) I might treat you as a good informant, again, if I know you would make a more educated guess on whether P than I would. (You know the company better than I and can make a better guess of whether its stock value will rise.) In this section, we will consider a stronger way in which one may treat someone as an informant. It is expressed in ordinary talk of “taking it” from persons. To take it from a person whether P is to treat her sincere word on whether P as settling for oneself the question of whether P. I take it from the coffee store proprietor that the shop is closed Sunday. I take it from an acquaintance in a casual conversation that she just got back from London. I do not take it from the person offering a tip on a stock that the stock will rise in value; at best I take it from her that there’s reason to think it might; or maybe that it’s more likely than not. Nor do I take from the weather forecaster that it will rain a week from today. In those cases, I do not take the relevant issue – whether P – to be settled , either in favor of P or in favor of ~P, by the informant’s sincere word.

To illustrate the informant-suggesting purpose I wish to discuss, consider once again a high-stakes bank case. Suppose, in such a case, you feel you can’t simply wait till Saturday to deposit the check. Enter Sally. If I say to you, “Oh, good, here’s Sally. She knows whether it’s open tomorrow,” I can thereby suggest Sally to you as an informant on the question of whether it’s open tomorrow – I suggest you can take it from her.

Notice here that we see some of the same patterns as before. In saying “Sally knows whether it’s open,” as a basis for suggesting you can take it from her whether it’s open, I’m making a little argument, “Sally knows, so you can take it from her.” If you deny my conclusion

(“no, you can’t take it from her”), you can respond to my argument in a number of ways. You could deny the premise that Sally knows whether it is open; you could claim that even if she knows, she is a liar, or won’t tell, etc, and so you can’t take it from her. What seems to be barred

13

is a certain kind of epistemic objection, this time centered on the informant: agreeing that the informant “knows whether P” but insisting that her grounds/reasons aren’t good enough for you to take it from her. This is in sharp contrast to a suggestion taking the form “Sally has a good idea whether it’s open” or even “Sally has excellent evidence concerning whether it is open.”

Those ways of recommending her as an informant do not close off the epistemic objection: there is nothing confused about the response, “yes, she has a good idea/excellent evidence about whether it’s open, but I can’t take it from her; her evidence just isn’t good enough (excellent enough).”

One might worry here, as before, that factivity is doing the work. If I say, “Sally has the right answer whether P; you can take it from her”, it seems you can’t properly raise the epistemic objection either. And the reason for this would seem to be the following. If the recipient agrees that Sally bears a factive cognitive relation to the question whether P, then the recipient must think that by believing what Sally believes on the matter the recipient would be believing truly on the matter. But this looks like a decisive reason to believe P if the informant says P and believe ~P if she says ~P, and so to take it from the informant whether P.

In order to set knowledge-attributions apart from attributions of other factive cognitive states with respect to informant-suggestion, we therefore should consider responses to suggestions of informants made by people other than the recipient, preferably in conversations not involving the recipient at all, so that there can be no issue of whether recipient gains evidence about the informant from the speaker.

Suppose you and I are watching a riveting film called The Bank Cases . In the film, the protagonist, Keith, needs an informant about whether the bank is open on Saturday. Suppose, also, that Keith needs quite good grounds before taking it as settled either way whether the bank

14

is open Saturday. His stakes are high, and there is a back-up “safe” option for him – just coming back tomorrow to deposit the check. Suppose I say about Sally, another character in the film,

“Sally knows whether it’s open Saturday, so Keith can take it from her.” Would it be misplaced and confused for you to raise the epistemic objection, to reply to me by saying, “Yes, Sally knows whether P, but Keith can’t take it from Sally because her evidence/grounds/reasons aren’t strong enough for him”? This does sound worse than the epistemic objection made against an ascription of “having a true belief about whether it’s open Saturday,” or “having a reasonable true belief.”

This data in favor of closing-off, admittedly, does not seem as clear-cut as in the case of agent-evaluation. However, I will take it as a working hypothesis that knowledge-attributing suggestions of informants close off this informant-centered epistemic objection, and that in this respect they differ from a host of other epistemic/psychological concepts, including those of true belief and reasonable true belief.

What is it about knowledge-attributions that allows them to serve the purpose of informant-recommendation in this way? A plausible hypothesis, as before, is in terms of reasons. One bit of simplifying terminology will be useful here, to avoid iterations of “warranted enough”: let’s say that an agent has P ( epistemically) available to her as a reason iff she is warranted enough in P for P to be a reason she has.

10

When I attribute knowledge that P to S in support of a claim that someone, S*, can take it from S concerning whether P, I’m treating S’s warrant concerning whether P as sufficient to make P (or not-P) epistemically available as a reason for S*. To accept my knowledge-attribution then commits one to doing so as well, and so to not raising the epistemic objection.

10 In what follows I’ll shorten ‘epistemically available’ simply to ‘available’.

15

What is it for S’s warrant concerning whether P to be sufficient to make P (not-P) available as a reason for S*? The idea would be that “transferring” S’s warrant for P (~P) to S* would suffice to make P (~P) available to S* as a reason. Or less metaphorically: if S* needs a certain level of warrant for P in order to have P available as a reason, S has that at least that level of warrant.

To summarize the two purposes:

The agent-evaluative purpose : where

-ing is an action or omission, by attributing knowledge that P to an agent, one can support one’s assessment of his

-ing in a way that closes off the epistemic objection.

The informant-recommendation purpose: by attributing knowledge whether P to a person S, one can support a claim that a recipient S* can take it from S whether P in a way that closes off the epistemic objection.

Knowledge-attributions can serve these respective purposes because, when they are made for these purposes, accepting them is to take on reason-related commitments :

In attributing knowledge that P to an agent as a basis for an assessment of his

-ing, where

-ing is an action or omission, one treats the agent as having P available as a reason to

(not to

), and anyone who accepts or agrees with one’s knowledge attribution commits herself, likewise, to treating the agent in this way.

In attributing knowledge whether P to a person S as a basis for a claim that a recipient S* can take it from S whether P, one treats S’s warrant on whether P as strong enough to make P (~P) available to S* as a reason, and anyone who accepts or agrees with one’s knowledge-whether attribution commits herself, likewise, to treating the informant this way.

§3. Implications for the contextualism debate

I have tried to describe these purposes and the reason-related commitments underlying them without prejudging the debate over the conditions of knowledge or truth-conditions of knowledge-attributions. The hope is that these purposes and commitments could be used as data in the adjudication of that debate. I now turn to that debate. I will ask whether and how the

16

concept of knowledge could have the features in question if one or another side in the contextualism debate is true. I begin with a brief account of the sides in the dispute (or at least the leading contenders):

11

Skeptical invariantism

– skeptical-standards invariantism.

Moderate invariantism – low- or moderate-standards invariantism.

Subject-sensitive invariantism (SSI) – takes the invariant conditions on knowledge-attributions to include a strong within-subject connection between knowledge and reasons (for action as well as belief). SSI is a form of fallibilism, and so allows knowledge to coexist with an epistemic deficiency that could make a difference in some possible practical environments. Thus, SSI holds that which standard one must meet to know P varies with one’s practical situation (the stakes, the available actions).

(Mainstream) contextualism

– affirms three claims:

(i) the standards that comprise truth-conditions for knowledge-attributions vary across contexts of attribution;

(ii) “internalism” about standard-determination; i.e., what determines the operative standard in any context is simply what beliefs, presuppositions and assumptions the speaker or conversational parties have.

(iii) in any context of attribution there is a single standard that comprises the truth- conditions for any knowledge-attribution.

A few clarifications on contextualism. First, talk of standards should be understood as

“intellectualist” in Stanley’s (2005) sense. That is, the standards do not discriminate between subjects with the same strength of epistemic position with respect to a given proposition. Thus, the sorts of conditions the SSIers take to be conditions of knowledge (e.g., being warranted enough for P to be a practical reason) will presumably not count as standards, since they can vary across subjects alike in these ways. Internalism about standard-determination, affirmed in (ii), in effect claims that speakers “call the shots” when it comes to which standards they are ascribing to subjects. How things stand with the subject, if the subject is not among the conversational parties, is irrelevant to which standards comprise the truth-conditions of knowledge-attributions

11 Noticeably missing is assessment relativism. See footnote 20 below for an argument that the claims I make about contextualism apply to relativism as well.

17

concerning him. (iii) has the welcome effect of guaranteeing the invariant truth of intellectualism about knowledge: “no difference in knowledge without a difference in strength of epistemic position.”

12

Mainstream contextualism is standards-based, internalist, and intellectualist.

In adjudicating the sides in this dispute, I will be concerned with questions of explanation: assuming such and such side is correct, would there be a good explanation of the two purposes of knowledge attribution defended above? Before looking at the details, I want to distinguish several kinds of explanation that might be given. I will assume throughout that explaining the purpose of a knowledge attribution amounts to explaining its associated reasonrelated commitment.

First, let’s say that an explanation gives a semantic vindication of a purpose if that explanation implies that when a knowledge-attribution is used for that purpose, the knowledgeattribution can be true only if the relevant reasons-facts hold. These reasons-facts will be those relevant to the reason-related commitments. So, in the case of agent-evaluation, a semantically vindicating explanation would imply that when a knowledge-attribution “S knows that P” is used for agent-evaluation, the knowledge-attribution can be true only if the agent has P available as a reason. Given a semantically vindicating explanation of the agent-evaluative purpose, we can see why it is that, when I evaluate Keith’s waiting till tomorrow to deposit the check by saying that he “knows the bank is open tomorrow,” your acceptance of this knowledge-attribution commits you to treating Keith has having the bank is open tomorrow available as a reason. It does so because the only way for that knowledge-attribution to be true is if Keith does have the bank is open tomorrow available as a reason. Mutatis mutandis for informant-suggestion.

12 This is a sloppy formulation, since it ignores the belief condition on knowledge. A better formulation is this: if

S1 and S2 have the same strength of epistemic position with respect to P, then either both or neither is positioned to know P.

18

The simplest way to give a semantic vindication is to take the relevant reasons-facts as required for knowledge itself. So, for instance, if one holds that knowledge that P requires having P available as a reason, one could say: when I defend Keith’s action by saying he

“knows P,” you are committed, by accepting my knowledge-attribution, to treating Keith as having P available as a reason because that’s something knowledge of P requires.

This simple – entailment-based – sort of semantic vindication isn’t the only kind, though.

This becomes clear once we consider contextualism. A contextualist such as DeRose (2009) does not take reason-facts to be entailed by knowledge, say, in the way he takes truth to be entailed by knowledge. Despite denying that knowledge entails availability as a reason, though, the contextualist might hope to explain how an attribution of knowledge – when made for the purpose of agent-evaluation – could be true only if the relevant reason-fact obtained.

13

The idea would be that, in any context in which it is used for agent-evaluation, ‘S knows P’ acquires a content which holds of S only if S has P available as a reason. This would be a semantic vindication but not of the simple entailment-based sort.

Another sort of vindication is pragmatic (as opposed to semantic). A pragmatic vindication is like a semantic one except that whereas the latter holds that the truth of the knowledge-attribution guarantees that the relevant reason-fact hold, the former holds that the truth of what is pragmatically imparted by the knowledge-attribution guarantees that the relevant reason-fact holds. It would be important here that to accept a knowledge-attribution carrying this implicature would itself amount to making that implicature oneself or at least endorsing what was implicated.

13 DeRose himself prefers to talk of “pressure” or “tendencies” to use standards appropriate to the purpose for which one is using the knowledge attribution.

19

There are other sorts of explanations one might provide which attempt neither a pragmatic nor semantic vindication. One might claim that the ordinary folk mistakenly think knowledge makes what is known available as a reason. Here I will be interested primarily with vindicating accounts, accounts which identify some proposition – either expressed by or merely pragmatically imparted by the knowledge-attribution used for the given purpose – which is true only if the relevant reason facts obtain. What makes such accounts vindicating is that they can show us what is wrong with violating the commitments, and so indirectly what force the commitments have. If by accepting the knowledge-attribution “Keith knew that P” I am accepting something which is true only if Keith has P available as a reason, and then I go on to treat Keith as not having P available as a reason, we can see that there is something wrong with me. Either something I’m accepting in accepting the knowledge attribution is false or I’m treating Keith as not having P available as a reason when in fact he does. To the extent that these facts are grasped by ordinary speakers, even if only implicitly and inchoately, we can see why it might well seem confused to raise the epistemic objection.

With this partial taxonomy of vindicating accounts in mind, let us turn to the sides in the contextualism dispute.

If knowing requires meeting the skeptical standards, then knowledge suffices for both within-subjects and across-subjects knowledge/reasons connections. For, if knowing guaranteed absolute epistemic certainty, then an agent’s knowing P would suffice to make P available to her as a reason; and an informant’s knowing whether P would suffice for her warrant to be strong enough, if transferred to a recipient, to make P available as a reason to that recipient. Skeptical invariantism thus gives simple semantic vindications of both purposes.

20

The only problem is that one’s knowledge-attributions would seem to be mostly false.

14

The little arguments “S knows the bank is open tomorrow, so S can just come back tomorrow” and “S knows whether the bank is open tomorrow, and so you can take it from S concerning the matter” would have false premises. They would thus not establish their conclusions, even if their conclusions were true. It would be like suggesting someone as a basketball center on the basis of the claim “he’s 7 feet tall,” when in fact he’s only 6 ½ feet tall.

One might say, though, that even though these attributions are literally false, still they are close enough to being true. So, although one’s argument, “S knows the bank is open tomorrow, so S can just come back tomorrow” fails to establish its conclusion because of its false premise, one manages to mean or impart an argument that does establish its conclusion, one having the form “S meets condition C with respect to the bank being open tomorrow, so S can just come back tomorrow.” Here C might be an anti-intellectualist condition (as per SSI) or it could be an epistemic standard selected in such a way as to guarantee that the argument’s conclusion is true if its premise is. The principal task of the skeptical invariantist is to explain how the “right” argument gets imparted when the semantics give us the “wrong” argument, and to explain this by appealing to general conversational principles rather than specialized principles about ‘knows’.

15

Moderate invariantism, by contrast, is unable to provide a semantic vindication of either of the two purposes. The hope is to provide a pragmatic one. A good pragmatic account, presumably, must appeal to relatively general features of conversational pragmatics, together with the semantics, to explain the postulated implicature. The best account I know of is due to

Brown (2006) and Rysiew (2007), who base their accounts on the maxim of relation (“be

14 It might be replied that even if this is so, a vast body of our belief does have warrant sufficient, in any practical environment, to make what is believed a reason. Think about

Obama didn’t live in the 7 th century, I’m not married to Aristotle, & etc. It might be better to call such a view “infallibilism” about knowledge rather than “skeptical invariantism.”

15 See Davis (2007) for an attempt to carry out this task, although Davis would not call his account a skeptical one.

21

relevant”). If the issue in a given conversation is whether the agent meets a very high standard, e.g., high enough to make P available as a reason to the agent, then even if the truth of a knowledge-attribution to the agent doesn’t guarantee any such thing, the knowledge-attribution, to be relevant, must communicate that the agent does meet this standard; and so by the maxim of relation we should expect there to be such an implicature. We could add to this account that when one accepts someone’s knowledge-attribution when made for one or other of the purposes

– by re-asserting it, saying “that’s true”, and the like – one makes that same implicature again or at least affirms it.

The difficulty I have with this approach is that it assumes the relevance of the knowledge-attribution and deduces that since the implied (relevant) proposition isn’t required for the truth of the knowledge-attribution it must be implicated or pragmatically imparted. What we need, by contrast, is an account of why the knowledge-attribution is relevant in the way it is. We need to be told why an attribution of knowledge – as opposed to various other attributions of epistemic status ( has good reason, etc.) – would speak decisively on this relevant matter. It is difficult to see how the explanation will go. By saying that a subject meets at least moderate standards, it is just not clear how or why this could address the key issue of whether the subject meets not only moderate but high standards.

The problem for moderate invariantism may even be worse than this. Given the widely accepted “assert the stronger rule,” moderate invariantism would not only fail to predict that claims to “know” have the asserted implicature; arguably, it would positively predict that they don’t. At least in a context in which the difference between meeting moderate and meeting very high standards was exactly at issue, by asserting that one meets moderate standards one would implicate, via the “assert the stronger rule,” that one does not meet high standards. The hearer

22

could work out the implicature as follows: “the issue is whether he meets high standards; clearly he meets at least moderate standards; so if he met the high standards, he would have said so; he didn’t; and so presumably he doesn’t.” 16

SSI is of course tailor-made to provide a simple semantic vindication of the agentevaluation purpose. It does this nicely. However, as DeRose has emphasized (2009, ch. 7), the going is not so smooth when it comes to the purpose of informant-suggestion. Here a semantic vindication is not possible. If what I’m claiming in saying “Sally knows whether P” is something like Sally has a true belief about whether P and Sally has warrant enough for P (or not-P) to make it a reason Sally has , then clearly my knowledge-attribution can be true even if the recipient needs more warrant than this to have P (or not-P) as a reason. Sally’s evidence is good enough for her to take it that P, but Keith needs more.

Thus, SSIers need to do some explaining. We need a story about why within-subjects knowledge/reasons connections should lead us to talk and think as if across-subjects knowledge/reasons connections hold. Here the SSIer must turn to a pragmatic vindication or some form of non-vindicating account. Again, the appeal to the maxim of relation would not seem promising. That the informant’s warrant is good enough for her just doesn’t seem relevant, at least not in the right way, to the question of whether it’s good enough for the recipient. It might be better to turn to a non-vindicating account.

17

I will not rehearse the arguments for and against these approaches, other than to second DeRose (2009, 234-8) in saying that the jury is still out and the prospects do not look as bright as SSIers might hope.

16 This criticism is the flip side of DeRose’s (2009, 118-24) criticism of Rysiew’s (2007) proposal about why it’s appropriate to say “I don’t know” in HIGH. I claim this may be a serious problem for moderate invariantism.

Whether it is depends on subtle issues concerning the best way to understand quantity implicatures.

17 See Hawthorne (2004), Stanley (2005, and Nagel (2010).

23

SSI takes the subject’s standards to “call the shots” and so it struggles with the informantrecommendation purpose even if it does fine on the agent-evaluation purpose. One might think that since mainstream contextualism takes the speaker’s standards to call the shots, it will have trouble with the agent-evaluation purpose even if it does fine on the informant-recommendation purpose. One might therefore fear a “big ugly tie” which might compel us to postulate an ugly ambiguity in ‘knows’. But DeRose (2009, 240) thinks there really is no tie here; contextualism wins hands-down:

“Fortunately, the big ugly tie is broken by the realization that, on contextualism, speakers’ own conversational interests can lead them to apply to subjects the standards that are appropriate to the practical situations faced by those subjects.”

So, the idea is that when we take certain conversational interests – agent-evaluation, in particular

– we can and should select standards appropriate to the subjects’ practical situation.

18

However, speakers can make mistakes about subjects’ practical situations. Suppose my wife, Frances, is aiming to evaluate my waiting in line for an hour to deposit the check. “Good lord, Matt, you knew you could deposit it tomorrow; you wasted your time. Suppose we are assessing Frances’ criticism, and she is not a party to our conversation. Now, if contextualism is true, then let’s ask how we could reject her argument’s conclusion, which we must do. We would argue as follows: what Frances said when she said Matt “knew” he could deposit the check tomorrow was true but due to the high stakes – higher than Frances realized -- Matt needed more evidence if he was to wait to deposit the check the next day. This is a version of the epistemic objection. We’re accepting the knowledge-attribution but rejecting the evaluation of the action on the grounds that more evidence was needed. If such contextualism is true, then acceptance of Frances’ knowledge-attribution does not commit one to treating Matt as having the bank is open tomorrow available as a reason.

18 See also Greco (2008)

24

The contextualist might reply that her account gives at least a partial semantic vindication

It does ensure this much: in any context in which an agent’s 

-ing is being evaluated by a speaker through an attribution of knowledge that P, if the speaker’s assumptions (or the

“common ground” assumptions in the conversation) about the agent’s practical situation are correct, then if the knowledge-attribution is true, then the agent had P available as a reason. So, anyone who accepts that knowledge-attribution and accepts the speakers’ assumptions about the agent is committed to treating the agent as having P available as a reason. This commitment disappears only when these assumptions are brought into doubt.

Still, if mainstream contextualism is right, the reason-related commitment does disappear under these conditions. But it doesn’t seem right to concede the truth of Frances’ knowledge attribution while insisting more evidence was needed. And Frances would presumably agree once she learns of her mistake about Matt’s stakes. Wouldn’t it be very natural for her to retract not merely her criticism of Matt but her knowledge-attribution? But that would be retracting something that is perfectly true. Compare the situation with having good/excellent reason to believe. If Frances instead asserted one of these in criticizing Matt, it would be perfectly fine, after learning of the actual stakes, to say, “I see, what I said about your having good reason to believe it was open were perfectly true, but you needed better reasons.” This is not fine, however, in the case of knowledge. Mainstream contextualism therefore does not give us a full semantic vindication of the agent-evaluative purpose. In order to do that, a theory would have to ensure that the truth of the knowledge-attribution that P guaranteed that the agent had P available as a reason, and it doesn’t guarantee this.

Similar points hold for the informant-recommendation purpose. Suppose I know Tom needs an informant on whether P. But I’m wrong about his stakes. They’re much higher than I

25

think. I say, “Sally knows whether P. Tom can take it from her.” Here, again, it seems the contextualist must say – at least assuming that the speaker’s assumptions about Tom’s stakes are determinative of the epistemic standard expressed by his use of ‘knows’ – that my knowledgeattribution was true but still Sally’s warrant for P isn’t good enough for P to be available to Tom as a reason. So, it ought to be fine for me, once apprised of Tom’s high stakes, to say, “what I said when I said Sally “knows” was true but she doesn’t have good enough evidence for Tom to rely on her.” That is, it ought to be possible to raise the epistemic objection using propositional anaphora like ‘what I said’ or ‘that’.

Here, again, the contextualist doesn’t give us all we want, but only this much: in any context in which a subject is being suggested as an informant concerning whether P for a thirdparty, if the speaker’s assumptions about the recipient’s practical situation are correct

, then if the knowledge-attribution is true, the informant’s warrant concerning whether P is good enough to make either P or not-P available as a reason for the recipient. The proviso concerning the correctness of the assumptions made in the speech context prevents the contextualist from giving a full semantic vindication.

Here I should also note the shortcomings of the sort of view Fantl and I have sympathetically explored in other work – (internalist) subject-sensitive contextualism. This view makes the availability of P to the agent as a reason an invariant condition of attributions to the agent of knowledge that P, but allows the content of such knowledge-attributions to vary across contexts of attribution insofar as when certain practical situations are salient it takes more to count as “knowing.” 19

It thus gives a full semantic vindication of the agent-evaluative purpose,

19 So, if S doesn’t have P available as a reason, then any attribution of knowledge that P to S is false; however if S does have P available as a reason, there might be some contexts which an attribution of knowledge that P to S is true and others in which an attribution of knowledge that P to S is false.

26

but because it relies on an internalist contextualist element it cannot do the same for informantsuggestion.

20

So far, I’ve only considered mainstream contextualism. On this sort of contextualism, if the speaker and her conversational parties all assume that a subject is in a low-stakes situation, and the purpose of the conversation is agent evaluation, the standards would be low, even if this assumption is false and the subject is in a high-stakes situation. But if we give up internalism about standard-determination, we can avoid this consequence. The guiding idea would be that the speaker manages to load into the truth-conditions of her utterance the standards that are in fact appropriate to the agent’s practical situation, not merely the standard that would be appropriate if the speaker’s assumptions about the agent’s practical situation are true. Thus, when Frances says “Matt knew that he could deposit it tomorrow; he wasted his time standing in line today,” her knowledge-attribution is true only if Matt meets the high standards appropriate to his situation, and not merely the low ones which Frances thinks are appropriate to his situation. A similar move is possible for the informant-suggesting purpose. When I’m wrong

20 Mainstream assessment relativism holds that knowledge-attributions vary in truth-value across contexts of assessment. Consider a “standards” implementation of this view, together with internalism about how the standards are determined. Call this mainstream assessment relativism. Can this view do better on the two functions of knowledge-attribution?

Recall the case of Frances’ criticism of Matt waiting in line: “Matt knew the bank was open tomorrow, so he should have just come back then.” Suppose, again, that Frances has falsely assumed that the stakes were relatively low. Now, if assessment relativism is true, then what Frances said is true relative to her context of assessment. So, isn’t the following a way to raise the “epistemic objection”? “What Frances said in saying Matt knew is true in her context of assessment, but Matt had insufficient evidence to do what he did.” The relativist might reply that in raising this objection, one isn’t accepting Frances’ knowledge-attribution and therefore the epistemic objection isn’t really being raised. To accept it – the relativist might say – Frances would have to either assert the proposition herself or use a non-relativized notion of truth in her response, saying “that’s true.”

This is an interesting response, but I think the commitment not to raise the epistemic objection can be understood to apply not just to acceptances as the relativist is understanding them but to “ok-ing” utterances or letting them pass. I certainly ok (in my North American winter) the Australian utterance ‘it’s summer’ and I even ok or let pass the speaking vulture’s utterance of ‘rotting flesh is tasty’. I refrain from challenging it. My arguments to the effect that knowledge-attributing evaluations of agents and recommendations of informants close off the epistemic objection, I think, show not merely that acceptance as the relativist understands it of the knowledgeattribution commits one to not raising concerns about insufficient warrant but that ok-ing or letting pass the knowledge-attribution does the same thing.

Finally, just as one can move to externalist versions of contextualism, one can move to externalist versions of assessment relativism, and I suspect, but will not explicitly argue, that the dialectic would be quite similar.

27

about your practical situation, and it’s much more urgent than I think, then I’m in fact applying high standards in asserting “Sally knows whether P,” even if I think I’m applying only low standards.

This proposal, which I’ll call direct-reference contextualism , seems to provide full semantic vindications of both purposes. When, in evaluating an agent, I assert “S knew that P,” what I say will be true only if S had P available as a reason, and so if you accept it, you commit yourself to treating S this way as well. When, in recommending an informant to S*, I assert to you, “S knows whether P,” what I say will be true only if S’s warrant for P (or not-P) is strong enough, if transferred, to make P (not-P) available to S* as a reason. How can this happen if the propositions expressed by the knowledge-attribution do not entail the relevant reasons-facts?

Even though there are possible worlds in which the proposition I actually express with my utterance of ‘S knew that P’ is true even though S does not have P available as a reason (worlds in which S’s stakes are much higher than they actually are, for instance), there is nevertheless a guarantee that the world of utterance is not among those worlds. To use Donnellan’s dthat operator, what I’m saying, in evaluating S’s 

-ing, when I say “S knew that P” might be represented as follows: “S meets with respect to P dthat -the standard S’s meeting of which with respect to P makes P available as a reason for S.” A similar account would apply in the case of informant-suggestion.

Unfortunately, this sort of contextualism abandons the raison d’être of contextualism, or at least the main motivating factor behind most flesh and blood contextualists: the upholding of intellectualism. The fact that the standards are determined externally and agent by agent will yield counterexamples to intellectualism. Suppose I’m evaluating two subjects, Mary and John, at once. I think, falsely, that they have the same stakes. I also think, falsely, that Mary’s

28

evidence is better than John’s. In fact, Mary and John have equally strong evidence but Mary’s practical situation is a humdrum low-stakes one whereas, unbeknownst to me, John’s is a highstakes one. I might claim, “Mary knows that P, but John doesn’t know that P.” What I’m really claiming is

Mary meets X and John doesn’t meet Y, where X and Y are determined by their actual practical situations. Given the actual facts, Mary does meet X, since X is a low standard, but John doesn’t meet Y, since Y is a very high standard. Thus, my claim “Mary knows that P but John doesn’t know that P” comes out true. A further claim is also true in my context, even if

I think it isn’t, viz. “Mary and John have the same strength of epistemic position with respect to

P.” It follows that in my context a sentence asserting a counterexample to intellectualism about knowledge is true.

So, the bad news is that the only view we have been able to find, short of skeptical invariantism, which provides a full semantic vindication of the two purposes cannot sustain intellectualism. Perhaps further amendments to the view could eliminate this, if we are willing to tolerate some ad hocery.

21

But I want to close with some speculative suggestions about which purposes need full semantic vindications and which do not.

21 John Greco suggested to me that the contextualist has another option I’ve not discussed. She could say, as Greco suggests in his (2012), that the attributor context picks out one practical environment, where a practical environment is defined by a set of relevant practical tasks, attaching, perhaps, to several actual and/or potential actors. Standards are determined by the practical environment so defined. In my case of Mary and John, since the purpose is agentevaluation, some single practical environment that one of them was in would serve to set the standards, and so we wouldn’t get the anti-intellectualist result I describe.

I have two concerns about this proposal (which really deserves a longer discussion than I can give here).

The first is that I don’t see what the mechanism would be to ensure that the “right” practical environment determines the standards. In a case in which the speaker thinks the practical environments of Mary and John involve only low stakes, and in which indeed Mary is in a low stakes environment, why would the standards be determined by the higher stakes environment of the other subject? I think there are also problems in explaining the mechanism whereby only one standard is determined when discussing the two subjects. Second, and here I go beyond things

I’ve argued in the body of the paper, I would argue that even in cases in which one isn’t attributing knowledge for the purposes of agent evaluation, attributing knowledge that P to a subject involves taking on a reason-related commitment concerning that she has P available as a reason. This is shown, I think, by the fact that it seems misplaced to assert, in reply to an attribution of knowledge that P to S, made for whatever purpose: “yes, what you said when you said S knows P is true, but S herself can’t rely on P as a reason, because she needs more evidence.”

It is misplaced, I think, not merely because it doesn’t address the speaker’s purpose. Admittedly, this claim is not as intuitively obvious as it is for knowledge-attributions made for the purpose of agent evaluation.

29

§4. Speculative Conclusion

The agent-evaluation and informant-suggestion purposes differ in one important way.

There is a coherent, and arguably even plausible, non-skeptical story to be told about how knowledge could in general require having what is known available as reason.

22

By contrast, there is no coherent non-skeptical story to be told about why, in general, one party’s knowledge that P should be good enough to make P available as a reason to any other party, however high her stakes. So, if skepticism is put aside, we should give up the hope to provide a simple – entailment-based – semantic vindication of the informant-suggestion purpose. But if skepticism is put aside, there may still be hope to provide such a vindication of the agent-evaluative purpose. What we need to do is see just how good a case can be made for the claim for the relevant within-subjects knowledge/reason connections. If a good case can be made, then certain views become less attractive. And here I mean not only moderate invariantism and mainstream contextualism but direct reference contextualism, too.

23

The problem is as follows. Suppose a subject S is in a high-stakes situation, and doesn’t have either P or not-P available as a reason. A speaker attributes to S knowledge whether P, suggesting her as an informant to a low-stakes recipient. Direct reference contextualism would make the knowledge-attribution come out true

(assuming the subject has a true belief whether P, etc). But then we have a knowledge-whether attribution “S knows whether P” which is counted true even though S has neither P nor not-P available as a reason.

24

There may be ways to modify direct reference contextualism to make it consistent with a within-subjects knowledge/reason connection. But if we are not skeptics, then to the extent that

22 See Fantl and McGrath (2009).

23 This includes Greco’s (2012) view as well.

24 I’m assuming that ‘S knows whether P’ is true in a context of attribution only if either ‘S knows P’ or ‘S knows not-P’ is true in that context.

30

there is a good case to be made for within-subjects knowledge/reason connections we should be prepared to think of the vindication of the informant-suggestion purpose as more of a patch-up job than the vindication of the agent-evaluative purpose.

25

Appendix

I have argued that knowledge-attributing evaluations of actions close off the epistemic objection. But we evaluate cognitive states by appealing to knowledge as well. If you are speculating whether Mary might come to our party, I might criticize you by saying, “You know

Mary can’t stand parties!” And if, looking out the window at the sun, you voice the belief “it’s going to be sunny today,” I can criticize this belief by saying, “you do know that the forecast was for early morning sun followed by rain all afternoon.”

Can we simply expand the range of evaluated states/activities to include cognitive ones?

A worry arises. If a kind of fallibilism about knowledge is true, one that allows knowledge to coexist with the absence of maximal epistemic certainty, then in principle there can be knowledge-attributing defenses of cognitive states which are vulnerable to the epistemic objection. Suppose I believe that I lived in South Carolina in 1976 – something true – with as much certainty as I believe that I have hands, with maximal psychological certainty. Suppose I defend my maximal certainty in this by saying “I know that I lived in South Carolina in 1976.”

This knowledge-attribution doesn’t seem to be a good defense, and that’s because it’s open to the epistemic objection, “yes, you know it, but you still shouldn’t be maximally certain; not as certain of it as you are of having hands or 2 and 2 being 4, because you’d need better grounds for thinking that you had lived in South Carolina then to be so certain of it.”

25 I would like to thank audiences at the University of Nebraska, Stirling University, the University of St. Andrews and the University of Oslo for helpful discussion, as well as David Henderson, Jonathan Ichikawa, John Turri, and especially John Greco for written comments on earlier drafts of the paper.

31

I think our discussion of reason-related commitments shows us how and why the epistemic objection isn’t closed off in this sort of case but is closed off in the cases I gave in the first paragraph of this appendix. In making the epistemic objection in latter examples, one would be failing to treat the subject as having the relevant proposition epistemically available as a reason, whereas this is not true of the former examples. Consider the sunny day example. To raise the epistemic objection against this criticism would be to say, “I know the forecast was for rain, but I’d need more evidence for thinking the forecast was for rain in order to abandon my belief that it will be sunny.” In raising this objection, one is not treating oneself as having the forecast was for rain epistemically available as a reason. Because if were so available, then one should give up one’s belief. Contrast the epistemic objection in the case of my defense of my extreme certainty that I lived in South Carolina in 1976. In replying, “You do know you lived in

South Carolina in 1976 but you’d need better grounds to be as certain of this as that 2+2-4,” you might still be treating me as having the known proposition epistemically available as a reason.

The key claim is that P is just not a reason to be absolutely certain that P. So, you can treat me as having I lived in South Carolina in 1976 epistemically available as a reason while also holding that I need better grounds for this proposition to be maximally certain of it. You can do this because I lived in South Carolina in 1976 isn’t a reason – let alone a justifying reason – to be maximally certain of its truth.

26

The difference between the cases, I think, is explained by fact that in defending one’s maximal certainty one is attributing knowledge that P as a direct evaluation of one’s maximal certainty rather than as an evaluation which depends on the fact that P would be a justifying reason to be maximally certain that P. In the case of action, it is not possible, I think, to give

26 Imagine arguing: well, Caesar conquered the Gauls, so that’s a reason we have to be maximally certain that

Caesar conquered the Gauls – indeed as certain as we are that 2 and 2 make 4.

32

these sorts of direct knowledge-attributing evaluations; the evaluation always depends on justifying relations between the thing said to be “known” and the action. In the cognitive case, direct evaluations are common and familiar. Consider the familiar criticisms: “it’s unreasonable for you to believe that P!” or “you don’t know that P!’ Consider the familiar defenses: “I am reasonable in believing P!” or “I know that P.” This difference helps us see why the epistemic objection can arise for cognitive

but cannot where

is an action or omission.

What can we say, in general, about the objections closed off by a knowledge-attributing evaluation of

-ing, where

-ing may be cognitive or practical, or really anything of the sort for which one can have reasons? I propose this: a knowledge-attributing evaluation closes off any objection, the making of which amounts to not treating the subject as having the “known” proposition epistemically available as a reason.

27

Works Cited

Brown, Jessica (2006). “Contextualism and Warranted Assertability Maneuvers,” Philosophical

Studies (130, 3: 407-435).

_____(2008a). “Subject-Sensitive Invariantism and the Knowledge Norm for Practical

Reasoning,” Noûs

42, 2: 167-189.

27 It seems there are some broadly practical

– plans for what to do in other sorts of situations – for which the epistemic objection isn’t closed off in the case of knowledge. Take 

to be planning to take gambles on P for tiny payoffs no matter what the penalty is if not-P should they arise. Isn’t the epistemic objection appropriate in response to a knowledge-attributing defense of having such a plan?

Suppose we understand the objections which knowledge-attributing evaluations close off to be objections which amount to violations of the commitment to treat the agent as having the “known” proposition available as a reason. The objection “you know that P, but you need better grounds for P in order to plan to take these gambles if they arise,” arguably does not fail to treat P available as a reason to the agent. P is just not a reason to plan to take such gambles if they arise. The explanation for this, in my view, is the same as the explanation for why P is not a reason to dismiss any possible counterevidence against P. For discussion, see appendix II of Fantl and McGrath

(2009).

33

______(2008b). “Knowledge and Practical Reason,” Philosophy Compass.

Cohen, Stewart (2004). “Knowledge, Assertion, and Practical Reasoning,” Philosophical Issues

14, 1: 482-491.

Craig, E.J. (1990). Knowledge and the State of Nature, OUP.

Wayne A. Davis (2007) “Knowledge Claims and Context: Loose Use,”

Philosophical Studies

132, 3: 395-438.

DeRose, Keith (2009). The Case for Contextualism, OUP.

Fantl, Jeremy and Matthew McGrath (2009). Knowledge in an Uncertain World, OUP.

_____ (forthcoming). “What Good are the Stakes-Shifting Cases?” in Brown, Jessica and

Adrian Haddock (eds) Knowledge Ascriptions, OUP.

Greco, John (2008). “What’s Wrong with Contextualism?” Philosophical Quarterly 58, 232:

416-36.

_____ (2012). ”A (Different) Virtue Epistemology,” Philosophy and

Phenomenological Research.

Hawthorne, John (2004). Knowledge and Lotteries, OUP.

Henderson, David (2009). “Motivated Contextualism,” Philosophical Studies 142, 1: 119-131.

Kvanvig, Jonathan (2003). The Value of Knowledge and the Pursuit of Understanding, CUP.

Nagel, Jennifer (2010). “Epistemic Anxiety and Adaptive Invariantism,” Philosophical

Perspectives. 24 (1):407-435.

Riggs, Wayne (2006). “The Value Turn in Epistemology,” New Waves in Epistemology , (eds.)

V. Hendricks & D. H. Pritchard (Ashgate).

Rysiew, Patrick (2007) “Speaking of Knowledge,”

Noûs

41 (4): 627–662.

Stanley, Jason (2005). Knowledge and Practical Interests, OUP.

34

Download