November 02, 2006

Contextualism vs. Interest-dependent Invariantism (Greco)

Thanks to Joe for the invitation to post.

I want to raise some considerations about attributor contextualism and interest-dependent invariantism. (These considerations come from a paper called “What’s Wrong with Contextualism” that I am working on for Philosophical Quarterly. ) First, I want to raise some methodological considerations about how we ought to decide the case between the two. Second, I want to argue that these considerations at least point us in the direction of attributor contextualism for knowledge attributions. Here is the gist of the argument: John Hawthorne and Jason Stanley have made a lot of the idea that knowledge is the norm of practical reasoning, using this plausible thesis to argue in favor of invariantism and against contextualism. But it seems to me that the argument can be turned on its head. In other words, the idea that knowledge is for practical reasoning seems to me to count in favor of contextualism. The main idea is this: If knowledge is for practical reasoning, then it is the relevant practical reasoning context that should govern. Sometimes this is the subject-context-- we make a knowledge attribution in the context of discussion about what S should do. But other times it is the attributor context-- we make the attribution while wondering what we should do. Still other times it is a third party's context-- we are discussing what some third guy should do. Technically this turns out to be a version of attributor contextualism, since the truth value of knowledge attributions varies across attributor context. The governing idea, though, is that truth values are relative to practical reasoning contexts.

Below this idea is developed more fully.

The usual methodology for adjudicating between contextualism and interest-dependent invariantism in epistemology is two-fold: consult our intuitions about possible cases, and consult the linguistic data regarding actual language use. We may include in the latter descriptions about how certain grammatical kinds in our language in fact behave. So, for example, it is common to describe the ways that indexicals behave, and to describe analogies or disanalogies with “knows.” This methodology invites different ways to explain the relevant data. For example, contextualists and invariantists will offer competing explanations for why a knowledge claim seems true or seems false in a possible case, with one side explaining the intuition in terms of semantic competence, the other in terms of pragmatics, perhaps together with “semantic blindness.”

I think we can supplement this two-fold methodology in a fruitful way. Specifically, we can ask what our concept of knowledge and our knowledge language are for. We can ask what roles they play in our conceptual economy and our linguistic practices. By doing so, I suggest, we gain further insight about how our concepts and language can be expected to behave. This same methodology has recently been defended by Edward Craig.

There seems to be no known language in which sentences using “know” do not find a comfortable and colloquial equivalent. The implication is that it answers to some very general needs of human life and thought, and it would surely be interesting to know which and how . . . .

Instead of begining with ordinary usage, we begin with an ordinary situation. We take some prima facie plausible hypothesis about what the concept of knowledge does for us, what its role in our life might be, and then ask what a concept having that role would be like, what conditions would govern its application. (2)


What might such a “prima facie plausible hypothesis” be? Here are two ideas that have received a lot of play lately: that knowledge is the norm of practical reasoning, and that an important function of our knowledge language is to flag good information and good sources of information. The first idea is emphasized by, among others, Hawthorne and Stanley. The latter idea has been defended in detail by Craig.

Human beings need true beliefs about their environment, beliefs that can serve to guide their actions to a successful outcome. That being so, they need sources of information that will lead them to believe truths . . . So any community may be presumed to have an interest in evaluating sources of information; and in connection with that interest certain concepts will be in use. The hypothesis I wish to try out is that the concept of knowledge is one of them. To put it briefly and roughly, the concept of knowledge is used to flag approved sources of information. (11)



Putting the two ideas together, we get the following plausible thesis: that an important function of our concept of knowledge and our knowledge language, perhaps its primary function, is to flag information and sources of information for use in practical reasoning.

Now suppose this is right. Does that speak in favor of attributor contextualism or interest-dependent invariantism? To my mind, it speaks in favor of attributor contextualism. More specifically, it speaks in favor of the version of attributor contextualism that allows the attributor context to be sensitive to the interests and purposes operative in the subject context. My thinking is this: if the function of knowledge is to serve practical reasoning, it should be tied to the interests and purposes that are relevant to the practical reasoner.

To make the point more clearly we may use Hawthorne’s notion of a “practical environment.” One’s practical environment is constituted by those aspects of one’s environment that are relevant to practical reasoning. Often enough, the practical reasoner with whom we are concerned will be in the attributor’s practical environment. Often enough, that is, one attributes knowledge for the purpose of practical reasoning in one’s own practical environment. But sometimes the practical reasoner will be outside the attributor’s practical environment. For example, sometimes we attribute knowledge for the purpose of practical reasoning in the subject’s practical environment. In that case, it would seem, it is the interests and purposes operative in the subject’s practical environment that are relevant.

These considerations suggest the following general rule: the truth-value of knowledge attributions (and the like) depends on the interests and purposes operative in the relevant practical reasoning context. Sometimes this will be the practical environment of attributor, sometimes that of the subject, and sometimes that of some third party. The position that results, however, will be a version of attributor contextualism, since it entails that the truth-value of knowledge claims is variable over attributor contexts. More exactly, the position is a version of interest-dependent, subject-sensitive contextualism.

As I said above, these remarks are at best suggestive. I don’t pretend to have established the present version of contextualism over its competitors. It is worth noting, however, that the proposed view does very well in relation to Hawthorne’s scorecard for evaluating contextualist and invariantist positions. In fact, it does better than any position that Hawthorne considers. Not pretending to have argued for these claims, I will simply assert the following: subject-sensitive contextualism respects the Moorean constraint that most of our knowledge claims are true, respects plausible closure principles, preserves the intuitive connections between knowledge, assertion and practical reasoning, and can (near enough) respect disquotational schemas for ‘knows’. We get this last result because all knowledge attributions must satisfy fairly high minimal standards, and so a knowledge claim in one context can normally be imported into another. I say “near enough” because there will be exceptions to this general rule. Specifically, we cannot disquote into contexts where stakes drive relevant standards unusually high. That there are such exceptions, however, seems correct. That is, we do not expect disquotation to go in that direction. (Hawthoren correctly notes that no anti-skeptical view can respect both the “Epistemic Possibility Constraint” (If the probability for S that p is not zero, then S does not know that not-p) and the “Objective Chance Principle” ( that epistemic probability follows knowledge of objective probability). (94))

Finally, the proposed view deals nicely with a kind of counter-example that gets posed against contextualism--ones involving attributions of knowledge to a high-stakes subject context from a low-stakes attributor context. For example, consider the case where we are considering whether S “knows” that the bank is open on Saturday, based on the evidence that he was at the bank two weeks ago and it was open on Saturday then. Nothing much depends on his being right for us but a lot depends on it for him. Intuitively, we should judge that S’s claim to “know” is false, even though we are evaluating his claim from a low-stakes context. As Hawthorne and Stanley point out, it seems that it is the interests and purposes operative in the subject’s practical environment that should govern the standards for “knowledge” here. But so long as the attributor context can be properly sensitive to the interests operative in the subject’s practical environment, attributor contextualism can accommodate this point. More specifically, insofar as it is the practical reasoning of the subject that is at issue in the case, the present view rules that it is the interests and purposes operative in the subject’s practical environment that ought to govern our evaluation of the knowledge claim. On the other hand, if the knowledge claim is being evaluated for use in our own practical reasoning, then it is the interest and purposes operative in our own practical environment that should govern. All that seems intuitively correct to me. That is, the proposed position seems to me to yield the right results in each case.

18 comments:

Jeremy Fantl said...

Hi John,

Nice post. But consider Stanley's "considerably less contentious principle that knowing that p is sufficient for acting on one's belief that p" -- his "far less controversial claim that if one knows that p, it is rationally permissible to act on one's belief that p." (These are endorsed in his replies to critics on his website.)

Whether the statement "it is rationally permissible for S to act on p" is true is not plausibly sensitive to the attributor's context. My stakes might be higher than S's, but if from S's context S is rational to act as if p, then I never speak truly when I say that it is not rational for S to act on p. Therefore, if Stanley's "far less controversial" claims are right, knowledge is sensitive to subject context.

Of course, this does not preclude that "knowledge" is also sensitive to attributor context. Interest-relative invariantism -- at least, to the extent that it says that two subjects with the same strength of epistemic position regarding p can vary with respect to whether they know that p, because of their different stakes in whether p -- is perfectly consistent with attributor contextualism.

Is this what you're endorsing when you endorse your subject-sensitive contextualism? If you are, then that seems fine. But, of course, we're still going to end up with many of the counterintuitive consequences of subject-sensitive invariantism: we'll still be left with the falsity of Stanley's "intellectualism" -- the position that knowledge is just a matter of properly epistemic features. And it will still be possible for S to be in a better epistemic position regarding p than S', even though S' knows that p while S does not.

-Jeremy

Aidan said...

I like the methodological point - Schaffer adopts a similar methodology in 'The Irrelevance of the Subject', and also suggests it favours contextualism over IRI. The idea that contextualism doesn't have to take the form of a thesis about the semantics of the word 'knows' seems to be gaining momentum right now, and so it looks like we're going to need suggestions for how to decide this debate that go beyond tests for whether that particular word is context-sensitve.

I just have a couple of remarks about the taxonomy you offer on p11.

I've been trying to persuade various people recently that there are invariantist interest-independent theories which are subject-senstive rather than absolutist. In fact, I take Hawthorne to sketch such a position in the first parts of chapter 4. There Hawthorne describes an invariantist view that makes whether a subject's true belief counts as knowledge sensitive to factors like what counterpossibilities are salient to that subject. But there's nothing interest-dependent about such a position as far as I can tell - it seems as intellectualist as versions of contextualism which make similarly heavy-play with saliency and the like. I think this suggests that Jason's label IRI is a more discriminating and useful label than SSI for his, Hawthorne's, and Fantl and McGrath's views (I see that you express a similar reservation about the label SSI underneath the table), and that the table perhaps oversimplifies things (even given its development on p13).

John Greco said...

Hi Jeremy. Thanks for your comments. You might be raising the possibility of a position I had not thought of-- one on which both a) knowledge is interest relative in the sense that SSI says it is, and b) knowledge attributions are context relative in the sense that attributor contextualism says they are. I mean to be defending the position that knowledge attributions are context relative in the sense that attributor contextualism holds, but that they are relative to practical reasoning contexts rather than attributor contexts. These can overlap, but it is the relevant practical reasoning context that governs, whether it is the same as the attributor context or not. I.e. the interests and purposes operative in the relevant practical reasoning context (which may or may not be the attributor's context) are what determine truth conditions for the knowledge attribution. Technically, this ends up being a form of attributor contextualism on the usual definition, since it entails that the truth values of knowledge attributions can vary across attributor contexts. But that is a consequence of the more fundamental claim of the position, which is that truth values are relative to relevant practical reasoning contexts.

As far as Stanley's "plausible principles" go, I am inclined to say the following. It will never be the case that one truly asserts "S knows p but it is not rational for S to act on the belief that p," since by talking about what it is rational for S to do we make S's practical reasoning environment the relevant one. Does that satisfy?

By the way, I plan to cite your stuff when I submit the final version.

John

John Greco said...

Aidan,

you wrote: "I've been trying to persuade various people recently that there are invariantist interest-independent theories which are subject-senstive rather than absolutist. . . Hawthorne describes an invariantist view that makes whether a subject's true belief counts as knowledge sensitive to factors like what counterpossibilities are salient to that subject. But there's nothing interest-dependent about such a position as far as I can tell"

I think you are right that room for such a view exists, but as far as I know, people who have invoked salience have thought that salience is largely determined by interests, so their views still make knowledge interest-dependent.

Jeremy Fantl said...

Hi John,

You say:

"It will never be the case that one truly asserts "S knows p but it is not rational for S to act on the belief that p," since by talking about what it is rational for S to do we make S's practical reasoning environment the relevant one. Does that satisfy?"

I think it gets me a bit clearer on your view, so in that sense it satisfies. But, one more bit of clarification. Will there be any cases in which it is not rational for S to act on p, but in which it is true to say of S that S knows that p? Here, we won't be asserting that S knows that p but S isn't rational to act on p. For, we may not know anything about S's stakes. But, given what S's stakes in fact are, S is not rational to act on p. Is it possible to say truly of S that S knows?

-Jeremy

Aidan said...

Oh yes, I wasn't claiming that anyone has actually defended such a view. But Schaffer presents pairs of cases in the paper I mentioned above which are designed to show that varying the stakes is actually a distraction in this debate. In his view, the real cases that favour contextualism don't involve change in either the subject or the attributer's practical enviroment; what changes between the cases in each pair is whether some counterpossibility is salient to the attributor ('one should consider unbiased minimal pairs that differ only in the salience of error': 89). The kind of invariantist view I had in mind would also be motived by such pairs of cases. But yeah, I don't know anyone who holds a view of this sort, and it doesn't strike me as very attractive. And if appeal to salience sneaks interest-dependence in through the backdoor, then it doesn't belong in the same box as absolutism, as I was in effect suggesting.

Jason Stanley said...

John,

There are a few worries with allowing lots of recherche facts to determine the semantic content of "know", relative to a context of use. One worry is that on your view, we don't know the semantic content of a knowledge attribution unless we know something about the practical situation of the subject of the knowledge attribution. In contrast, according to we anti-intellectualists (Hawthorne, Fantl and McGrath, and me), ignorance of facts about the practical environment of the subject may lead to ignorance about the *truth-value* of the knowledge attribution, but not ignorance of its semantic content. I think our result is the better one. If someone says of Hannah that she knows that the supermarket is open now, surely I know what is being said of Hannah, even though I may not know whether or not it is true.

Brit Brogaard said...

That's a neat objection but doesn't the same problem arise on any theory that treats a given non-purely indexical expression (or purely indexical expression for that matter) as context-sensitive? Consider the following objection to treating quantified noun phrases as containing implicit domain variables: "If you say 'John was flunked by every professor' surely I know what is being said of John, even though I may not know whether or not it is true".

Jason Stanley said...

Brit,

Yes, of course that is the strategy used by Cappelen and Lepore throughout their book to argue for semantic minimalism. But it is not a good general strategy. For the alternative view to context-sensitivity about domain restriction is that 'John was flunked by every professor' expresses a false statement about every professor in the universe. This is a manifestly implausible thing to take as what I said. Furthermore, it's pretty clear (especially given that the alternative minimalist claim) that you *don't* know I said unless you know the domain restriction (of course, there are problems of vagueness and slight differences, which Cappelen and Lepore exploit, but set those aside). If I am speaking of every professor at Rutgers, and you think I am speaking of every professor at Michigan, you haven't understood what I said. If Greco's suggested version of contextualism were right, then, like in the case of domain restriction, you would need to know the practical facts to know what was said. But it doesn't seem there is a corresponding case like the 'Rutgers professor-Michigan professor' case for domain restriction for knowledge attributions. Furthermore, the alternative non-contextualist theory (IRI) explains in a satisfactory and non-error theoretic way the proposition that is expressed (unlike in the case of minimalism about quantifier domain restriction).

Brit Brogaard said...

Hi Jason,
Yes, I like the restriction strategy much better than the error-theoretic approach. But suppose someone could come up with an alternative non-error theoretic approach to quantification which takes the semantic content of quantified noun phrases to be constant across context (just like IRI for the case of 'know'). In that case it seems you would need to concede that the alternative would be the better theory.

Jason Stanley said...

Brit,

Sure. But such a theory for quantifier domain restriction is unimaginable.

John Greco said...

Aidan,

It looks like I should read that Schaffer paper. Thank for the reference.

John Greco said...

Jason and Brit,

Thanks very much for your comments. Does the following passage (now added to my paper) help with Jason's concerns?

Stewart Cohen points out that attributor contextualism need not hold that “knowledge” refers to different properties in different contexts. Rather, it is open to the contextualist to hold that “knows” always picks out the same n-place relation, with at least one of the relata being fixed by attributor context. For example, the contextualist can hold that “knows” picks out a three-place relation between a person, a proposition, and an evidential standard, with attributor context determining the value of the standard place. In that case “knows” picks out the same relation, and therefore the same property, across all attributor contexts. In this sense attributor contextualism is on a par with interest-dependent invariantism. In both cases, the idea is that knowledge is a many-place relation, the relata of which are partly fixed by context. The contextualist has attributor context doing the job, whereas the invariantist has subject context doing the job. In neither case need we say that “knowledge” picks out different properties in different contexts.

John Greco said...

Jeremy,

"you wrote: one more bit of clarification. Will there be any cases in which it is not rational for S to act on p, but in which it is true to say of S that S knows that p? Here, we won't be asserting that S knows that p but S isn't rational to act on p. For, we may not know anything about S's stakes. But, given what S's stakes in fact are, S is not rational to act on p. Is it possible to say truly of S that S knows?"

Yes, on my view there will be such cases. For example, suppose we are concerned with our own practical reasoning context, where stakes are low. Relative to our purposes, it might be true to say that S knows p, and rational for us to act on p. But it might also be that S's practical reasoning context is very different, and such that it is not rational for S to act on p.

Brit Brogaard said...

Hi John
Contextualism and IRI are on a par with respect to the character of 'know'. And since IRI takes the character of 'know' to simply be the content, there is a sense in which contextualism and IRI are on a par with respect to understanding (but not with respect to understanding what was said).

Keith DeRose said...

John: Your idea sounds in the same ballpark as what I propose in section 7 of my "The Ordinary Language Basis for Contextualism and the New Invariantism" (Philosophical Quarterly, 2005). I wonder how they seem to you to relate? Is it the same idea? The paper is available at

http://pantheon.yale.edu/~kd47/OLP.htm

but here are a couple of relevant paragraphs:

Fortunately, that is not the situation we face, for contextualism can in fact easily handle the second type of case. What is vital to see here is that there is nothing in contextualism to prevent a speaker's context from selecting epistemic standards appropriate to the subject's practical situation, even when the subject being discussed is no party to the speaker's conversation – which is good, because speakers often do select such standards when their conversational purposes call for it. On contextualism, the speaker's context does always call the shots. Hawthorne is right about that. But sometimes speakers' own conversational purposes call for employing standards that are appropriate to the practical situation of the far-away subjects they are discussing, and so the shot that the speakers' context calls can be, and often quite naturally will be, to invoke the standards appropriate to the practical situation faced by the subject being discussed. And one type of conversational situation in which it will be quite natural for speakers to (and perhaps even strange for them not to) employ standards appropriate to subject's situation are contexts in which the speakers are discussing whether or not the subject "knows" in connection with evaluating a practical decision (including, it's worth adding, decisions concerning whether to assert something) that the subject faces (or faced, will face, could have faced, etc.): "She should do A only if she knows that P", "Well, if he knew that P, he could have responsibly done A", "She should assert that P only if she knows that P." In such a context, where whether the subject "knows" is being discussed for such a purpose, it seems strange indeed to employ epistemic standards not appropriate to the practical situation faced by the subject.

But, as is equally evident, and as we have already seen, there are other, very different contexts in which we have quite different reasons for discussing whether a subject "knows," and in which it is perfectly appropriate to, and in which speakers actually do, apply to subjects standards quite different from those that the subject's own practical situation calls for (and not because the speakers are mistaken about the nature of the subject's context). And, disastrously, SSI cannot handle this evident fact, since SSI has the subject's situation set the standards that govern any speaker's description of that subject, whatever the speaker's context.

John Greco said...

Hi Keith. Thanks for those comments. I have read your paper, which I found helpfulful, as usual. One important relation between your ideas and the ones in my paper is I think this: I am trying to provide a principled reason for thinking that standards (and other context-sensitive features) are determined by speaker context and/or subject context, depending on the case. The reason: knowledge attributions are governed by the relevant practical reasoning environment, and this, in turn, is because knowledge is for practical reasoning.

My paper is concerned to do some other things as well, though. One is to explore whether a virtue-theoretic approach (agent reliabilism, if you like)is committed to attributor contextualism. I originally thought yes. Now I conclude no, but that contextualism is nevertheless the more attractive option, for the reasons I am saying.

Finally, my paper draws a conclusion that I assume you will not like. I argue that considerations regarding the relationship between knowledge and practical reasoning suggest that, although standards will vary across practical reasoning contexts, they will not vary widely. But this sort of stability threatens the anti-skeptical moves that contextualists typically want to make. My conclusion is that contextualism is correct as a semantic thesis, but the considerations that show this undermine its effectiveness as an answer to skepticism.

More specifically, if knowledge is for the purpose of practical reasoning, this requires that knowledge be widely attainable, and this puts pressure on the standards for knowledge in a downward direction. The argument here is straightforward: Knowledge is required for practical reasoning, and practical reasoning is widely necessary. Therefore, knowledge must be widely available to serve its purpose. Therefore, the standards required for knowledge cannot be so high as to make knowledge less than widely available. That is, the standards cannot be so high if knowledge is to have the role that it does in our practical activities.

(Likewise, the demands of practical reasoning require that knowledge be imported across various practical contexts, and this puts pressure on the standards for knowledge in an upward direction. Suppose that I need to know whether p, and that in a different context S claims, “I know that p.” For S to be a useful informant, for her to be a source of knowledge to be used in my practical reasoning, I have to be confident that the standards by which S counts as knowing in her context are at least as high as my practical context requires. Again, the information-sharing function of our knowledge language puts pressure on the standards for knowledge in an upward direction.)

I know that in recent papers you have made your contextualist response to skepticism more subtle. Perhaps you can agree with everything I say about the stability of standards, but still invoke contextualist mechanics to explain our skeptical intuitions
in philosophical contexts. The intutions will turn out false on this view, but a natural mistake. Does that seem right?

In a final version of the paper I will cite you (and Cohen) in the way you suggest. The present version still contains some confusions that I need to weed out, so if you are interested in reading the paper at some point, please wait a few days for a new version to be posted.

123 123 said...
This comment has been removed by a blog administrator.