Blog Coordinator

X-Phi Grad Programs

« New CFP Site | Main | Armchair in Flames? »



Feed You can follow this conversation by subscribing to the comment feed for this post.

Jennifer Nagel

Interesting paper. I do think it makes a large difference to have the subject of your vignettes assert of herself that she has knowledge in each case. And I'd agree that this does a lot to reduce the variation in her apparent confidence that may be present in Stanley's versions -- a variation that, as you note, makes it more difficult for him to claim that he has isolated the impact of stakes on knowledge ascriptions.

But on his behalf -- by having the subject declare that she knows in each case you might be undercutting an important feature of the cases as he presents them to his philosophical audience; namely, the stipulation that in each vignette the subject has the same non-practical basis for her claim. A charitable reader might take Hannah's self-ascription of *knowledge* after she is challenged to implicate that she possesses some further evidence to rule out Sarah's concern about the possibility of a change in hours. (After all, we'd ordinarily expect a high-stakes subject to have done a bit more homework on whether the bank really would be open the next day; perhaps your subjects are guided in part by that expectation in their assessment of Hannah's claims to know... .) So I think perhaps more should be done to ensure that your subjects are judging the parallel cases under Stanley's stipulation that precisely the same evidence is being relied upon in each case. (And to get your case really tight, you might want to ensure that all the truth-relevant features of Hannah's belief formation are the same across the shift in stakes -- not only does she have to have the same evidence, she also has to be processing this evidence in the same way.)

Incidentally, did you test Ignorant High Stakes cases? Just curious.

Josh May

Thanks for the comments, Jennifer!

I think you're right that there is a worry that Stanley could press about having the protagonist make the self-ascription of knowledge, saying "I know that the bank will be open." It's true that stating "I know that p" in conversation with someone often expresses that one has a certain confidence that might convey to the other person that one has more evidence for the truth of p than one has revealed---or something along those lines. Stanley could then say that this additional hidden evidence (that subjects in the studies are assuming Hannah has) prevents the high practical interests from yielding something less than knowledge.

Schaffer has suggested to me a similar idea as well---except as a point in his favor as opposed to Stanley's. He suggested that Hannah's self-ascription of knowledge in our cases may prevent the alternative from being salient or accepted as a live option.

So there are things along such lines that both Stanley and Schaffer may be able to say. But for this study at least, we wanted our vignettes to be as close to Stanley's originals as possible. And Stanley does have his protagonist make self-ascriptions of knowledge. And our main goal was to see if people say what he says about the two High and Low Stakes cases from the book and to test whether Schaffer was right in his critique of those Stanley on those specific cases. But maybe it would be worthwhile to test vignettes that address such issues about self-ascriptions of knowledge. (Maybe we or someone else will do that in the future!)

However, I doubt that further studies would get the differences needed for either camp (i.e. Stanley or Schaffer), because:

(1) Regarding Stanley: Adam Feltz and Chris Zarpentine produced data like ours from cases (their bridge cases) that don't seem to have such problems. Those cases seem to avoid the problems because: (a) the cases don't involve the protagonist asserting "I know" to another person (he only thinks it to himself), (b) the cases seem to make it clear that the evidence on which the protagonist is basing his thought that he knows is the same in both low and high stakes cases, and (c) they seem to make it clear that he processes the evidence the same as well. But their cases don't address the concern with respect to alternatives, only stakes.

(2) However, regarding Schaffer: Wesley Buckwalter's study on contextualism and the bank cases doesn't have the protagonist say "I know..." in any of the cases, including the one in which the possibility of error is raised. Thus, there shouldn't be a problem with any self-ascription removing the salience of the alternative. Yet Buckwalter still found similar data showing that there is no difference in the judgments made about the relevant cases.

So, it seems like the move suggested is doubtful. But I'm sure more studies are warranted to test these and similar issues.

About Ignorant High Stakes: we didn't test it. However, Feltz and Zarpentine did run all of Stanley's original cases, and the results for all of them were not in Stanley's favor.

Geoff Pynn

I don't understand your explanation (in fn 3) for why you don't follow Stanley's example of having the speaker say "I know" in one context and "I don't know" in another and then asking subjects whether they think the speaker has spoken truly. Like Jennifer I think think having the speaker say that she knows makes a big difference. (First-person report: when I consider HS-A I think something like, "well, she must know something Sarah doesn't about how *this* bank isn't going to change its hours!")

It seems to me that regarding briefly-described cases where a speaker's assertion is not obviously incompatible with anything in the case (as in all of your cases), we're *strongly* inclined to regard her as speaking truly. So I'm not at all surprised at your findings -- a subject who reports disagreement with "The speaker knows p" would probably regard the speaker in the case as speaking falsely. But I don't see what they tell us about contextualism or SSI.

The entire "intuitive" case for contextualism and much of the case for SSI rests on the supposed fact that it seems that a speaker can truly assert "I know p" in one context but truly assert "I don't know p" in another context when all that's changed about the cases are the practical importance of p's truth and / or what alternatives are salient. It would be very interesting to find out experimentally whether this really *was* a fact, but that's not what you've done.

Your data would "indicate that Stanley's view of common intuition is incorrect," as you say, if the following premise were true:

(X) If subjects agree with the statement "P" concerning a hypothetical case where the subject asserts "P", then, of a case identical in every way except that the speaker asserts "Not-P" instead, subjects would not regard the speaker as speaking truly.

But for the reason I suggest above (and others!) I find (X) to be not at all plausible!

(None of this should be taken as disagreeing with you when you say that your results are "quite interesting" -- they are!)

Josh May

Thanks for the comments, Geoff! Two comments/responses:

(1) Our footnote 3 is not intended as an explanation of why we didn't follow Stanley in asking people whether the protagonist says something true. It's only an attempt to justify our modifying Stanley's original HS case in a way that has the protagonist claim to know instead of not---i.e. to diffuse the objection that our cases are now biased because of that choice.

The following, however, is an attempt at an explanation of why we didn't follow Stanley in asking people whether the person says something true: it's more natural, I think, to ask people whether they agree with something, instead of whether some person says something true. We could have e.g. said: "Hannah says something true when she says that she knows.... How much do you agree or disagree?" But that just seems unnecessarily awkward and unnatural. And naturalness is, I think, highly important in vignettes that are supposed to elicit ordinary judgments.

Regardless, though, Feltz & Zarpentine did follow Stanley in asking whether what the protagonist says is true, but the results are the same. So it doesn't seem to matter which way it's phrased.

(2) You may be right that (X) is false. But I don't think we need it to make the point we want to make. We might need principle (X) if our target was any general argument for anti-intellectualism that is based on its being common sense or part of ordinary linguistic practice. But our goal, with respect to Stanley, was just to test whether what Stanley says *about the cases from his book* is what most ordinary people would say---which is something Stanley seems to claim. Stanley may very well be right that most people would provide the anti-intellectualist judgments about *some or other* pair of low versus high stakes cases. But he takes the cases from his book to be good candidates for this. We wanted to see if they are.

When we say that our data "indicate that Stanley's view of common intuition is incorrect," we do not mean that he is wrong that interest-relative invariantism (IRI) is intuitive. Stanley doesn't exactly say that. He says that the intuitions he has about the *specific cases from his book* are ordinary, intuitive pieces of common sense. That's the claim about ordinary intuitions that we think our data indicate is incorrect. The only thing more that we claim is the following: The truth of this first claim of ours (i.e. that Stanley is wrong about what our common sense intuitions about such cases are) casts some doubt on his argument for IRI *from the book*---i.e. the argument that says IRI best explains the judgments he says we have about those cases. It can't be a point in favor of IRI that it explains such-and-such asymmetric judgments about the cases from the book if we don't have the alleged asymmetric judgments/intuitions. (While this may seem like a fairly weak claim, we still think it's substantial, especially given how influential the bank cases have been.)


Hi Josh,

Thanks for your response. I don't mean to be too nit-picky, but I just want to clarify my concern.

At issue (I'm assuming) is whether Stanley's conjecture in this passage is correct:

"Here, I take it, are the intuitive reactions we have about these cases. In Low Stakes, our reaction is that Hannah is right; her utterance of 'I know the bank will be open' is true. In High Stakes, our reaction is that Hannah is also right. Her utterance of 'I don't know that the bank will be open' is true." (KPI, 5)

These are the supposed data that motivate IRI. If your study gives us reason to doubt these data, then I think that would be an extremely significant result -- because it would not only undermine Stanley's case for IRI, but the cases for contextualism and relativism about 'knows' as well.

But I don't think you've given us reason to doubt these data. Your study confirms Stanley's observation about Low Stakes. And you haven't directly tested his claim about High Stakes. You've tested a different (possibly related!) claim that he could have made instead: to wit, that if the case were just like his High Stakes case except that Hannah said "I know that the bank will be open" instead of "I guess you're right. I don't know that the bank will be open", we would regard her claim as being *false*. The fact that your subjects tend to agree that Hannah knows in HS-A gives us good reason to be suspicious about *this* claim. But Stanley doesn't make it, and I don't see how he's committed to it.

Is the assumption that if subjects agree with what Hannah says in HS-A, then "the intuitive reaction" to Stanley's High Stakes case is not what he says it is? I just don't know why we should believe this.

(PS I am quite sympathetic to the difficulty posed by the unnaturalness of asking subjects how much they agree with 'S truly said p'. I still have a problem with the methodology, primarily because if it's controversial whether one of the terms in the relevant sentence is context-sensitive -- as it is here -- then not figuring out a way to distinguish use from mention in the test is going to be a serious difficulty. But that's another issue for another time.)

(PPS - I'd be eager to see the Feltz / Zarpetine data you mention -- if it uses Stanley's original High Stakes case, complete with Hannah's denial that she knows, and gets results at odds with Stanley's claim, then I take back almost everything I said...)

Josh May

Hi, Geoff.

You're right that we don't test exactly what Stanley says, given that we don't have Hannah deny herself knowledge and then ask subjects about whether she says something true. So I think you're right that one could try to make the move of bringing out your principle (X) and holding that it's false. However, three responses:

(1) I think Stanley wouldn't want to make that move. I don't think I would either, if I were him. That is, I think Stanley would agree with, as you say, our "assumption that if subjects agree with what Hannah says in HS-A, then "the intuitive reaction" to Stanley's High Stakes case is not what he says it is [in the book]." Maybe he would have other qualms with our studies, but I doubt it would be along the lines of denying principle (X).

(2) And I think it was warranted for us to change Stanley's HS case in the way we did in order to avoid the problem of making Hannah seem like she fails to believe or have sufficient confidence in the truth of the proposition in question. Although, I guess the findings of Feltz and Zarpentine show that we needn't have made the change---people don't tend to judge that Hannah doesn't know even if she says she doesn't.

(Note: Does that show that the folk don't think that belief or a certain level of confidence is required for knowledge? Not necessarily. People could just be thinking that she does still *believe* it, etc. Perhaps they figure that Hannah denied herself knowledge only because she thinks Sarah's point shows that she doesn't have enough *justification*, etc.)

(3) Besides, I do think the Feltz and Zarpentine studies suggest that this all isn't an issue anyway. They use Stanley's exact scenarios, so they have Hannah say in HS that "I guess you're right. I don't know that the bank will be open." And they then ask the subjects whether what she says is true. The results are strikingly similar to ours. (The differences in their data---which aren't much anyway---could just be from the problems with Stanley's original cases.) I would email Adam Feltz to ask for a draft if you'd like to check it out.

Thanks again for the detailed comments. I appreciate the nit-picky-ness!

Dorette van der Tholen

Dear Josh May,
I just read the latest draft of this article which was very interesting. I noticed that you conducted an analysis of variance with your data. Do you have some info on the suitability of doing this test with your sample, especially, did you have a normal distribution for the target question?

Josh May

Hi, Dorette.
I'm afraid I'm not the one to talk to about the stats. (I wish I knew more than I do, and I plan on acquiring that knowledge at some point!) We had psychologist Jay G. Hull on board to help with that sort of thing. I'm sure he wouldn't have us run the ANOVA if it wasn't appropriate for the sample size. But I'm not so sure about the normalcy of the distribution. It was certainly enough to make it reasonable to suppose there wasn't a real effect. But I'm not sure I know enough about normal distribution to say whether it meets the criteria strictly. I'll ask him about some of these details and get back to you.

By the way, what do you have in mind for the target question? Are you wondering about the distribution for each cell in each experiment, or just one in particular?

Josh May

Hi, Dorette.

It looks like the idea is this: with any kind of reasonably sized sample (like the ones we had), the distribution of the means associated with the statistical tests can be assumed to be normal and that doesn't depend on the distribution of the original observations. This assumption is a standard one given the Central Limit Theorem.

I hope that answers your question(s). Thanks for the interest in our paper!

The comments to this entry are closed.