Friday, November 23, 2007

C&R Zeman: A Closet Contextualist?

According to David Lewis (1980), a context is a location (spatiotemporally centered world) where a sentence may be said (but need not contain any utterance nor speaker at the center etc.), and thus has countless features, and an index is an n-tuple of shiftable features of context. Moderate views have it that a sentence s is true at a context c iff s is true at c with respect to the index of that context i_c; and radical relativist views such as MacFarlane's depart from that.

With respect to this framework, one can characterize contextualist versions of moderate relativism endorsing the appearances of sentence s being true at c (wrt i_c) while false at c* (wrt i_c*); and in turn one can distinguish indexical contextualism (having it that this is true in virtue of s having a different content at c than c*) from non-indexical contextualism (having it that s has the same content at c and c* but that determines a different value wrt i_c than wrt i_c*.

Contexts in this sense are very rich. In particular, there is nothing as the epistemic situation (or standard or whathaveyou) of the context. There is that of the speaker at the center of the context (if one), that of the attributee of the utterance at the center of the context (if one), that which is salient in the conversation that takes place near the center of the context (if one), and so on and so forth. As Dan Z points out, this richness of contexts tends to be neglected in some discussions about knowledge attributions, and more sophisticated versions of indexical contextualism would presumably exploit this. (He still thinks that the view suffers from other “quite serious” difficulties so that it is “likely” that it will fail. I’m not convinced, but let’s discuss that in some other occasion.)

As I understand his own positive proposal, he claims that the attributions have the same semantic value across context, but are evaluated differently with respect to different indices of these context—where the epistemic standard of the context that figures as a coordinate in the index need not be that of the subject at the center of the context, nor the attributee, but is the highest (I guess among those that are relevant in the conversation that takes place near the center). But thus his seems to me to be a version of non-indexical contextualism and not radical relativism proper!

7 comments:

Dan Zeman said...

There are many notions that I employ in my paper that are far from being reasonably clear, and the notion of context is doubtlessly one of them. Often, I fail to specify which kind of context I’m speaking about (of use? of assessment?), and I don’t even bother to give at least a working definition of context, such as that provided by Dan L. This unclearness is, I think, the root of Dan’s worry that my view, although loudly and proudly declared relativistic, might be of more moderate sort – and thus, if I got the implicature right, less interesting.

I will try to defend the interestingness of my view by showing that the view is compatible with the relativistic framework. To begin with, I wasn’t very preoccupied with specifying where exactly my position lies in the taxonomy I provided in the paper. Now, Dan has provided another taxonomy, which partly overlaps with that of mine. (I will stick to Dan’s terminology, to ensure uniformity.) That is, in the paper I didn’t really considered non-indexical contextualism (NIC) as a separate option, because by setting the stage in the way I did (by assuming that the attributor of a K-claim is the speaker), NIC would have come to make the same predictions and have the same problems as relativism (given another assumption that I will disclose shortly). Thus I considered only indexical contextualism (which I simply call “contextualism” in the paper), for which the relevant standard in establishing the truth-value of a K-claim is that of the speaker, subject-sensitive invariantism (SSI), for which the relevant standard is that of the subject, and relativism, for which the relevant standard is that of the assessor. (In relativism’s case, the assumption I made was that the assessor = the attributor, so the view comes down to saying that the relevant standards is that of the attributor.) NIC would thus fall somewhere between indexical contextualism and SSI, in that the relevant standard for establishing the truth-value of a K-claim would still be that of the speaker, but it’s relevance would be of a different sort (as being part of the circumstance of evaluation rather than of the content).

Claims about the radicalness of a certain view are quite popular in philosophy. It is sensible to say, though, that the degree of radicalness one is after depends on where one draws the line in a given taxonomy. Given the taxonomy employed here, for me the important line to be drawn is that between indexical contextualism and all the other positions. The intuitive difference here is related to two roles context has, which could be stressed by using MacFarlane’s terminology: for indexical contextualism context has a content-determinative role, whereas for the other positions it has a circumstance-determinative role (this is true for relativism as well: what the context of assessment does is to select the “right” standards (that of the assessor) that go into the circumstances with respect to which a K-claim, as uttered in a given context of use, is to be evaluated). [I do actually believe that these two roles are two fundamental aspects of how context works, that could be distinguished in some principled way (think of Perry’s Z-landers), but that’s too complicated an issue to get in.]

This is not meant to be an excuse for my unclearness, though, so I should say something more. I think my view could be easily conjoined with the relativistic framework. Here is how MacFarlane defines truth in “The Assessment-Sensitivity of Knowledge Attributions”:

A sentence S is true at a context of use CU and context of assessment CA just in case for some proposition p,
1. S expresses p at CU and CA, and
2. p is true when evaluated at the circumstances determined by CU and CA.

On the view MacFarlane calls “propositional relativism” (which is the one I’m interested in here), sentence S expresses the same proposition across contexts of use – proposition that in turn is assessment-sensitive (its truth varies with the context of assessment). What is crucial here is the notion of context of assessment. Now, my view per se doesn’t add any formal complexity to relativism: it is just a way of understanding the relativist framework (especially the notion of context of assessment) and making it work. (What I do is giving practical advice, as it were.) As Dan remarked, I adopt DeRose’s insight that context is “very rich” – that it should be open as to comprise the relevant standard, even if it’s not that of the speaker, and even if the subject is not located nearby the speaker, etc. For DeRose, of course, the issue matters as to what should go into the content of K-claims. But why couldn’t one apply the idea to a view according to which context has a circumstance-determinative role, and not a content-determinative one, such as relativism? What DeRose does for the context of use, I do for the context of assessment. It should become “richer”, such as to comprise the relevant standard, even if it’s not that of the assessor, etc. In this way, I claim, cases that are problematic for the other views (indexical contextualism, NIC, SSI, unmodified relativism) could be easily handled (IF one agrees with the intuitions I claimed we have in those cases – which is, to put it mildly, an utterly controversial point). The modified relativist view will still have it that K-claims are assessment-sensitive, but the relevant circumstance that gets selected for evaluating a given K-claim is not necessarily that of the assessor.

[I’m not sure about this, but I think the view is also compatible with NIC, with a corresponding modification concerning indices. Truth would be defined as Lewis does in “Index, Context, and Content” (“sentence s is true at context c iff s is true at c at the index of the context c”), but the index of the context would have to be such that to comprise the relevant standard, even if it’s not strictly that of the context (in Lewis’ sense of the term “context”). Thus the only modification should concern epistemic standards, in that the world and time parameters are those of the context (again, in Lewis’ sense), but the epistemic standard parameter need not be. Again, I’m not sure you could really do this in the Lewisian framework – although Nikola Kompa has recently made some claims that would qualify her as NIC who employs a “richer” index of the context.]

Thus, I’m not giving up relativism, I’m just supplementing it with elements that make it better fit to cope with troublesome cases.

Dan López de Sa said...

Oh, maybe I misunderstood you in the paper. I thought your view was that the relevant "epistemic standard" that figures as a coordinate in the index need not be that of the subject at the center of the context, nor the attributee, etc, but was anyway provided by the context ("of use"), so that you respected that a sentence is true at a context c iff it is true at context c at the index of that context i_c.

If you allow the relevant coordinate of indices to be filled by elements of perspectives ("contexts of assesment")--locations from where a given sentence as said in a (possibly dfferent) context is viewed or assessed--, then (but only then) the view would indeed depart from moderation.

Do you agree?

Dan López de Sa said...

I was thinking about this yesterday and now I feel a little confused. For according to your proposal the epistemic standard that figures as a coordinate in the index is the highest, but the highest among which? I kind of assumed it was among those that are relevant in the conversation that takes place near the center of the context. But this supports the non-indexical contextualist version of your view. Which is the relevant class if you turn radical relativist?

Dan Zeman said...

The highest epistemic standard is to be chosen between that at the context of the subject and that at the context of the assesor/artributor (I made the simplifying assumption that the assessor is the same person as the attributor - I discuss the harmlessness of this assumption in the paper, footnotes 13 and 15). This will make it easy to account for cases like Low Attributor (or Assessor) - High Subject, in which we get the counterintuitive result that a positive K-claim, made by the attributor/assessor to the efect that the subject knows, is true. (To me it seems that the K-claim is false - but this is something that many people refuse to accept, MacFarlane included.) Since the attributor/assessor and the subject are the most relevant "characters" in the way in which I presented the cases, it is their epistemic standards that matter (but there might be some other standards in the vicinity that will matter in some situations - I haven't thought of that). Thus, if you agree that this is a counterintuitive result, then relativism is in trouble. Now, if you allow the attributor/assessor to choose as the relevant epistemic standard that of the subject, the problem is solved. Allowing that is still compatible with allowing that "the relevant coordinate of indices to be filled by elements of perspectives", so the view counts as relativistic. It's just that the "perspecives" need to be broader, so to speak.
As for the compatibility with NIC, again, I'm not sure that it works, since on my view you will have to accept that the epistemic standards that are relevant to the evaluation of a given K-claim should go beyound that of the concrete situation in which the K-claim is produced. This is probably something you cannot get from the Lewisian framework without seriously changing it. But if so, nothing is lost for my proposal.

Dan López de Sa said...

Sorry I'm still confused.

The highest epistemic standard is to be chosen between that at the context of the subject and that at the context of the assesor/artributor (I made the simplifying assumption that the assessor is the same person as the attributor - I discuss the harmlessness of this assumption in the paper, footnotes 13 and 15).

If contexts are locations (spatiotemporally centered worlds) where a sentence may be said, then both the speaker of an attribution and the subject who is its attributee "are" in the same context. And if you require the "assessor" of the sentence at a context to be the attributor, and thus something provided by the context ("of use"), then the view turns again to be non-indexical contextualism as opposed to radical relativism. Which might be all for the good, of course :-).

Dan Zeman said...

I'm not sure I can be clearer than before, but here are some points:
1. I never committed myself to Lewis' notion of context.
2. There is a clear sense in which two people having different standards (whatever that means: difference in what is at stake for them, difference in what and how many epistemic possibilities they need to rule out, etc.) are said to be in different contexts. If, however, one doesn't want to use "context" to mark the difference, one is free to use other terms - "situation", for example. But saying that "both the speaker of an attribution and the subject who is its attributee "are" in the same context" would erase the difference. Cases like LA-HS and HA-LS wouldn't make sense and, I guess, neither will the original contextualist cases (L and H) in which the same attributor/subject is in different situations (i.e., has different epistemic standards).
3. As I said, equating the assessor with the attributor is just a simplification (in order to be in line with the way I framed the cases, in which feature just an attributor and a subject). To see the full potential of the view, consider the following case. Person A is the assessor of a K-claim made by person B, in which knowledge is attributed to person C. Now, A is in a situation in which epistemic standards are low, whereas C is in a situation in which epistemic standards are high (B's situation doesn't matter). Acording to relativism, A would assess the positive K-claim "C knows that p" as true. However, I take this to be a wrong result (based on the intuition I have that C doesn't know, given her high standards). According to my view, A should take C's epistemic standards as relevant in assessing the K-claim, and thus yielding the right result. (This raises the problem of the availability for A of C's epistemic standards, for it is obvious that in some cases we cannot just glance into others' situation and get information. I discuss solutions to this problem at the end of the paper.)

Dan López de Sa said...

Ad 1, I state things with respect to the Lewisian framework because I think it is clear, detailed, and useful. Of course I'm more than happy to try to rephrase the points in any alternative framework you would favor. Any suggestion?

Ad 2, there is a clear sense in which two people have different "standards," to be sure, but this difference is simply not erased by saying that "both the speaker of an attribution and the subject who is its attributee "are" in the same context", why would it?

Ad 3, if you allow the relevant coordinate of indices to be filled by elements of perspectives--the "standard" of the assessor, regardless of whether she is the attributor of the utterance or not--, then the view does indeed depart from moderation. So if this is in effect the case in your view, then I agree it is certainly not (closet) contextualism ;-)!