Andy Egan has posted a new draft of his paper on “Epistemic Modals, Relativism and Assertion”, first noted here in September. I re-read it over the weekend, especially since Thony Gillies and I are working on a response to the various relativistic analyses of epistemic modals. Our paper is still nascent, although I appear to have committed myself to presenting it in Germany in a month from now, so it better coalesce until then. I thought I would expedite some of my thinking on the topic by posting a couple of questions about it.
Question #1:
Andy presents the following scenario:
James Bond has just returned to London after a long day of infiltrating SPECTRE’s secret base in the Swiss alps, planting a bug in the main conference room, and slipping out by night after leaving persuasive but misleading evidence of his presence in Zürich. Sipping martinis in MI6 headquarters while monitoring the newly placed bug, Bond and his CIA colleague Felix Leiter overhear a conversation between Blofeld and his second in command, Number 2.
In the course of a discussion of the pros and cons of various nefarious plans, Number 2 says to Blofeld, Bond might be in Zürich.
[BTW. In my quotes, I have corrected the inexplicable absence of the umlaut in "Zürich" in Andy's paper.]
Andy’s discussion proceeds to report the following intuitions: (i) Number 2’s utterance is appropriate given his state of evidence, (ii) Blofeld is in the same basic state of evidence and so can easily agree with Number 2 by saying You’re right or the like, (iii) Leiter on the other hand (sipping martinis next to Bond in London) couldn’t really say That’s right or the like.
In the first draft of the paper, Andy had also claimed that it is clear that Leiter could and in fact should say That’s false!. One deduces from the changes in the new draft that there was some pushback on this issue. In a footnote (Fn. 5 on p. 5f), Andy says that some people seem to resist having Leiter say That’s false! and proposes an explanation along these lines: saying That’s false! could be taken as an accusation of Number 2, that Number 2 made some mistake, and people don’t want that. [I assume that it is to be assumed that Number 2 and Blofeld are blamelessly misled by Bond's planted evidence -- otherwise it should be OK to accuse them of an error.]
In the same footnote, Andy claims that when we consciously avoid an interpretation in terms of blaming Number 2 for an error and “when we think of Felix’s attribution as indicating a firm refusal to agree, we have no such hesitation”.
I actually don’t at all share that intuition. To my ear, Leiter couldn’t really say to Bond I don’t agree when they hear Number 2 say Bond might be in Zürich. That sounds, if possible, even weirder to me than That’s false!.
My quick intuition is that if Leiter were to say I don’t agree, he would thereby be inserting himself into the conversation between Number 2 and Blofeld, something that he can’t really do.
Do people agree with my assessment?
Question #2:
Andy wants to defend a relativistic semantics for epistemic modals. Basically, might p is said to be true for i in w at t iff p is compatible with the evidence within i’s reach in w at t.
From this, it is predicted that Bond might be in Zürich is true for Number 2 but false for Leiter.
Andy then goes on to worry about the general picture of communication and what role relativistic propositions could possibly play in conversation. He proposes that one can assert a relativistic proposition only if there is a presupposition that the proposition has the same value for all the participants in the conversation.
Here’s my point: given that proposal, it would seem that Leiter cannot properly insert himself into the conversation between Number 2 and Blofeld. After all, it is clearly not presupposed by anyone that Leiter and Number 2/Blofeld have the same evidence in their reach.
So, it would seem that Andy’s proposal actually predicts that Leiter should not be able to say I don’t agree. As I indicated, I think that’s the right prediction to make. But it is of course in conflict with the way that the data are presented, especially since the data are supposed to give prima facie reasons to explore a relativistic semantics.
My question in a nutshell then: if we actually take Andy’s analysis of communication with relativistic propositions and apply it to the Number 2/Leiter scenario, don’t we predict different judgments than the ones Andy reports?
In the end, Thony and I are working towards a demonstration that the apparently relativistic data presented by Egan et.al. and MacFarlane can be treated within a rather standard non-relativistic semantics for epistemic modals. It seems to me that with Andy’s analysis of communication, the relativistic story becomes virtually indistinguishable from a non-relativistic story. But arguing for that will have to wait for another occasion. For now, I just wanted to raise the two points above.
I’m confused about the example. Why would Leiter say anything to Bond about where Bond is, given that it’s common ground between them that they’re in MI6 headquarters?
Perhaps a better example (for drawing out “I disagree” intuitions) would involve two groups of spies. Group 1 is spying on Group 2, and neither group knows where Bond is. A spy in Group 2 says “Bond might be in Zürich,” and the Group 1 spies hear this. One spy in Group 1 believes that Bond has left persuasive but misleading evidence that Bond is Zürich, and she believes that Bond is in fact not in Zürich. But none of this is common ground within Group 1. So on hearing what the Group 2 spy says, the Group 1 spy rightly says “I disagree,” (or even, perhaps, “That’s false”) and goes on to explain to her Group 1 compatriots why she thinks Bond isn’t in Zürich. This is consistent, I think, with one of the other Group 2 spies rightly saying “I agree” (in response to the initial “Bond might be in Zürich”). I think that’s all Andy wanted.
November 15th, 2004, at 5:24 pm #Eric, thanks.
I do agree that the Group 1 spy can say “I don’t agree”, just like someone in Group 2 could, who had some doubts about the reliability of the evidence that Bond is in Zürich.
Is it clear what presuppositions are in place with respect to everyone having the same epistemic reach?
November 15th, 2004, at 6:58 pm #I think you’re right that the members of Group 1 don’t presuppose that their “epistemic reach” is the same as Group 2’s. Perhaps Andy’s idea is that the Group 2 spy who says “Bond might be in Zürich” must presuppose that he and his addressees have (roughly?) the same epistemic reach. When the Group 1 spy says “I don’t agree,” he must presuppose that he and his addressees have roughly the same epistemic reach, too. But that’s okay.
If this is what Andy wants to say, I’m a bit confused about how Group 1 is supposed to interpret the Group 2 spy’s assertion. If I say to you “This is fragile” I presuppose that you’ll be able to figure out what I’m referring to with ‘this’. But if a spy who doesn’t know what I’ve demonstrated overhears us, she won’t be able to figure out what proposition I expressed. I don’t see why Group 1 isn’t in a similar situation with respect to Group 2. Group 1 knows that Group 2 all presuppose that the members of Group 2 have the same epistemic reach. But Group 1 may not know whether their epistemic reach is anything like Group 2’s. In what sense can they disagree, then, with the Group 2 spy’s claim that Bond might be in Zürich?
November 15th, 2004, at 11:40 pm #Hi Kai and Eric - sorry to come in on this late.
Kai, thinks for the post, and for fixing the umlauts.
Eric, thanks - the case in your first post is exactly the sort of thing I’m after. The Bond/Leiter/SPECTRE case was supposed to be relevantly similar.
(I think what went funny was that both of my group 1 spies had the relevant evidence for Bond’s not being in Zurich. So while my case gets the inappropriateness of “that’s true”, it doesn’t get the appropriateness of “that’s false”, ‘cuz there’s no good communicative purpose being served by Felix saying something about his and Bond’s evidential situation that Bond already knows. Does that seem like the right diagnosis?)
So if that’s right, then I think I’m happy - the revised case gets appropriateness of “that’s false”, and then we’ve got our motivation for going relativist. (And I think I’d want to say that my original case gets the truth of “that’s false”, but there’s funniness about assertability because it’s already part of Bond & Felix’s common ground.)
Eric, I’m confused about the worry in your second post.
The spy in group 2 (call him Number 2) asserts the CW proposition that’s true in all and only the triples s.t. it’s compatible with everything in i’s epistemic reach (i.e., everything that i knows, or could find out by one of the contextually relevant methods) at t in w that Bond is in Zurich.
Since it’s compatible with everything that’s in Number 2’s epistemic reach now in @ that Bond is in Zurich, what he’s asserted is true relative to him. Since his buddiies in group 2 all presuppose that the same things are within their epistemic reach, his assertion is okay.
Now take the group 1 spy (call him Felix). He understands English (which is what everybody important speaks in Bond movies…), so he knows, in the usual, tacit way that we know such things, that Number 2 expressed the CW proposition that’s true relative to a triple iff it’s compatible with what’s within i’s epistemic reach at t in w that Bond is in Zurich. It’s not compatible with everything that’s in the Felix’s epistemic reach now in @ that Bond is in Zurich, so what Number 2 said is false relative to Felix.
What else does Felix need to know in order to figure out what Number 2 said?
There’s definitely something funny about disagreement with CW contents - it doesn’t seem rght to say that, for all self-locating P, whenever you believe P and I believe not-P, we disagree. But that’s not the same problem as kot being able to figure out which proposition got expressed.
Re: question 2 - The picture wasn’t supposed to be that Felix was introducing himself into the other conversation. What it’s appropriate for him to assert just depends on what’s common ground between him and Bond. It’s supposed to be a case of commenting, as part of one conversation, on something somebody said in some other, non-overlapping conversation. So there needn’t be any presupposition that Leiter, Blofeld, and Number 2 are in relevantly similar epistemic situations in order for Leiter’s disagreement to be appropriate. There’s only such a requirement if Blofeld and Number 2 are part of the audience, which they’re not.
(Like I mentioned above, though, Eric’s case is better than mine for getting appropriateness of Leiter’s disagreement in his conversation, because in Eric’s case Leiter’s disagreement conveys some information about the group 1 spies’ situation that wasn’t already presupposed.)
December 4th, 2004, at 2:51 am #Hi Andy,
I think your diagnosis of what went wrong in your initial case is close but not quite right. There is often
good communicative purpose'' served by saying something that your addressee already knows---even if you know that your addressee already knows it. For example, I might say to my sonYou ate all the cookies” even though I know that he knows that he ate them, just because I want him to know that I know that he ate them. So, as I suggested earlier, I would explain the inappropriateness by (1) observing that it’s already common ground between Felix and Bond that they’re in MI6 and (2) citing Stalnaker 1978.Thanks for your response to my second post. I think I understand your view better now, so let me say some things about it, and you can tell me whether I go wrong. Here are (what I take to be) two facts about sentences that include indexicals: (1) What is common ground helps determine their assertibility conditions; (2) What is common ground helps determine the conditions for fully understanding particular utterances of them. (By `fully understand’ I mean something like this: you fully understand my saying “It’s pretty here” if you know where I am; you don’t fully understand it if you don’t.) You’re committed to (1) for epistemic modals, but you deny (2). Number 2’s assertion of ZURICH eliminates from Group 2’s context set the non-w,t,i triples centered on Group 2’s epistemic reach. Because Group 1 overhears the assertion, it also eliminates from Group 1’s context set the non-w,t,i triples centered on Group 1’s epistemic reach. And both groups fully understand Number 2’s assertion.
If all this is okay, then I have a new(ish) question. Assume that no one in Group 2 demurs when Number 2 asserts ZURICH, and that Group 1 knows this. What triples are eliminated from Group 1’s context set concerning Group 2’s context set? How does this work?
Also, does Group 2 know what their epistemic reach is? They can’t, since though they presuppose that they have a common epistemic reach, what their epistemic reach is is at best merely distributed knowledge. But then, how do they `know how to update’ on the assertion of ZURICH?
December 4th, 2004, at 9:59 pm #Yeah, that’s definitely right about the assertability stuff - I wasn’t distinguishing in the way I should have between what both parties knew and what was common ground.
Pre- the two questions, let me just say what I’m claiming happens with self-locating assertion, to make sure we’re on the same page. (I’m gonna call group 2 ‘SPECTRE’ and group 1 ‘MI6′, since I have some neurological condition that prevents me from keeping the 1s and 2s straight.)
Here’s what Number 2’s assertion does to the SPECTRE context set: it eliminates all of the triples s.t. the evidence within i’s epistemic reach at t in w rules out Bond’s being in Zurich.
Before the assertion of ZURICH, let’s suppose, the SPECTRE context set includes some possible situations whose occupants have within their reach conclusive evidence that Bond is in Tahiti (or London, or whatever). After the assertion, it doesn’t - it only contains CWs s.t. it’s compatible with i’s evidence at t in w that Bond is in Zurich. So a bunch of possible evidential situations have been ruled out.
What happens to the MI6 group when they hear this depends on a bunch of things. They know what proposition Number 2 asserted (let’s call it ZURICH, since we don’t have italics). But they don’t change their context set by eliminating all of the triples that aren’t members of Zurich. That’d be bad, since it would leave them presupposing that their evidence doesn’t rule out Bond’s being in Zurich, which is wrong - there he is with the martini.
Still, they learn something from overhearing the SPECTRE exchange. It’s just not ZURICH. Instead, what they learn is that that the members of SPECTRE all presuppose ZURICH.
But that’s just the same as what eavesdroppers learn from overhearing stuff in general - what an eavesdropper who understands what’s been asserted learns from an uncontested assertion is something about what the people he’s listening in on presuppose. For whatever proposition P got asserted, what the eavesdropper learns is that the eavesdropp-ees (now) presuppose P. If there’s a bunch of eavesdroppers, and the output from the bug is conspicuous enough, as in the MI6/SPECTRE case, the MI6 group will come to presuppose that the SPECTRE group presupposes P. So far, no difference between self-locating and other kinds of assertion.
If he takes the people he’s listening in on to be well-informed, he’ll be able to draw some more conclusions. This is where the self-locating (SL) and non-self-locating (NSL) cases come apart. In the NSL case, the eavesdropper gets to conclude that P is true. In the SL case, what they get to conclude is that P is true relative to the parties to the conversation they’re listening in on.
(So in the MI6 case, what gets eliminated, when they overhear Number 2’s uncontested assertion of ZURICH, is all of the triples that are not listening in on some conversation, all of the parties to which are ones of which ZURICH is true. That’s the answer to question 1, I think.)
On question 2: On my story, EMs are supposed to be making assertions about epistemic reach - so it’s actually pretty important that people don’t know what their epistemic reach is. What you need, in order to update in the right way, is just to know which so-far-un-ruled-out possibilities about your epistemic reach are incompatible with what’s being asserted.
I’m afraid I may have gone incoherent in there somewhere, but I think it’s better to just post than to try and edit…
December 5th, 2004, at 9:18 pm #Thanks, Andy—that helps a lot. I misunderstood your view in some fundamental ways but I think I see it better now.
You say that “EMs are supposed to be making assertions about epistemic reach - so it’s actually pretty important that people don’t know what their epistemic reach is.” Right. If people did know everything there is to know about what their epistemic reach is, then on your view there wouldn’t be much need for EMs in the first place.
But if a speaker is to meet any of the plausible sounding epistemic/doxastic norms of assertion when she uses an EM, she has to know or believe some things about her epistemic reach (and thus, on your view, some things about the epistemic reach of the people she’s talking with). Focus on the “Assert only what you believe” norm for now; norms that are harder to meet only make my question more pressing. If I can’t find my car keys, and my wife says “They might be in the kitchen,” then as I read your view she’s met this norm of assertion only if she believes that it’s consistent with what I know that they’re in the kitchen. But this is setting the bar way too high. She might have no opinion about whether I’ve even looked in the kitchen, and just be suggesting that, if I haven’t already looked there, then I should look in the kitchen.
I use examples like this one to help motivate my speech-act view, so I’d be interested to hear what you think of them.
December 6th, 2004, at 2:34 pm #I’m not sure how much trouble the car key case is for me. Suppose you’re rummaging for keys, and your wife says “they might be in the kitchen”. If you’ve already gone over the kitchen with a fine-toothed comb, you should dissent, by saying something like, “No, I already looked there”. On my view, what’s happened is that you’ve dissented from her assertion, refusing to accept that it’s compatible with your evidence that the keys are in the kitchen, and informing her that it wasn’t really compatible with her evidence, either. (Using “x’s evidence” to mean “what’s within x’s epistemic reach”.) That actually seems like the right kind of account of what’s going on to me.
There is some lingering funniness about the appropriateness of assertion there though, isn’t there? It seems like it’s inappropriate to say “they might be in the kitchen” if she thinks that you have looked there, and looked thoroughly enough to rule out that that’s where they are. But it does seem like it’s still appropriate if she doesn’t have a view about whether you’ve looked or not, even though it’s appropriate for you to dissent, and for her to retract, if it turns out that you have looked. Hm.
The kinds of phenomena that you talk about in your paper (I haven’t read the new draft yet - still in Sydney) that make me really worried about my view are the cases that look like adding-to-context-set effects, where you use “might p” to “rule in”, as it were (so to speak, if you know what I mean) some possibilities that had previously been ruled out. I have a vague idea about how I can explain that use of EMs, but I’m not very confident about it.
December 6th, 2004, at 7:17 pm #The only trouble I had in mind with the car key case was that last bit at the end of your second paragraph. Your content oriented explanation in the first paragraph seems plausible enough. But I don’t see how, on your account, she could meet the epistemic norms for assertion without believing that I hadn’t looked in the kitchen. But she doesn’t have to believe this to appropriately use the EM. (There are substantive norms for EMs, of course, but I’m not going to try to spell them out here.)
It’ll be good to talk about the deeper worry in the third paragraph sometime soon.
Oh, if you do come across a new draft of my paper on this, please save me some work and send me a copy! (You’re thinking of a new draft of a different paper.)
Thanks, Andy.
December 6th, 2004, at 8:52 pm #I shall say answer question #1, but. firstly, sorry if I do not respond to all comments.
The situation is quite common when people are watching TV: what is shown is quite different from what the viewers know to be the case. People judge these cases false but do not utter comments like That is false with these exact words. They usually correct the presupposition, by commenting on what they see or hear.
One common concrete comparable scenario: a movie shows some scenes that supposedly take place in Rio de Janeiro, but all details are wrong. The characters speak Spanish and drink tequila, and everyone is celebrating Mardi Gras in July and wear Caribbean styled costumes, until one of them falls into the Amazon and goes down the Iguazu falls (like in one of Bond’s movies by the way). Well, for most parts viewers often laugh at the awful mistakes, but when the Amazon river appears in the middle of the botanical garden in Rio the usual reaction is The Amazon is in the extreme Northwest, Rio is in the Southeast, silly! or They do not know geography! and, after the Iguazu falls cause more laughters, they would complete Iguazu falls are in the South, thousands of miles from the Amazon! or This error is even more grotesque.
I guess in the case you mention, Leiter would utter the words: Bond is here, you morrons! or something like that. Of course, the correction of the premises eventually imply that what is said or shown is false, but the sentence that is false is almost never the exact sequence of words used.
Those are my intuitions. But I do not know if this is attached to the use of a modal like might. It is not necessary that someone says You might fall into the Amazon, when in Rio for the audience to react against it correcting the presupposition. If one says You will fall into the Amazon, when in Rio it causes the same reaction. It is not even necessary to say anything, it suffices to show something out of order to cause such comments.
March 2nd, 2005, at 2:37 pm #