July 12, 2004

Testimonial knowledge from multiple sources

How about that-- a massive thread on the epistemology of testimony, focusing on this case (presented by Jon Kvanvig):

You have multiple testimonial sources for p. None of them have knowledge, since none of them have a total body of evidence sufficient for epistemic justification. But each of them has some evidence for the claim, which is why each of them tells you that p is true. They are all correct: p is true. And you believe them, so you have a true belief. Moreover, if you combine all of their reasons for believing p, the combination is a secure, ungettiered epistemic justification for believing p. So if the quality of their reasons is somehow behind their epistemic authority, can the multiplicity of sources cover over the epistemic flaw that each of them displays in reporting to you something they donít know to be true? In short, even though they donít know, do the details about their grounds for belief allow for you to know on the basis of these multiple testimonial sources?

As you may know, I don't think knowledge is particularly important compared to justification, but my immediate answer is: Hells yeah. This even seems like a relatively sound method of doing history--if you have a bunch of imperfect observers, each of them may be likely to make an error, but they're not all likely to make exactly the same error. So a belief based on many imperfect witnesses may be properly very robust, which I think is characteristic of most knowledge.

But, as comes out in Jon's comments, the question is whether this is really pure testimonial knowledge, in which one simply accepts the speaker's word for what is said, or is inference involved? I think it's hard to pick out exactly when inference is involved, but I could imagine Jon's case working as followed. The first person says p, and you think, "Maybe." The second person says p, and you think "Probably." The third person says p, and you think, "Lookin' good." The fourth person says p, and you think "Definitely." This seems pretty non-inferential to me--as you get each new report, you simply come to believe them more and more.

I think this matches Sandy Goldberg's condition (iii) in this comment (as amended in this one) that pure testimonial knowledge "depends for its status as knowledge on and on nothing more than, (a) the reliability (or safety or sensitivity) of the testimony and (b) the hearerís epistemic right to rely on that testimony." The hearer in this case is depending only on the (imperfect) reliability of the many testifiers. There is no individual piece of testimony for which her knowledge satisfies both (a) and (b), but it satisfies both (a) and (b) when you take the total testimony on which she's basing her belief.

(Note that I also hold that, in a case in which S tells you that p, and you think "Is S trustworthy? Yes" before forming the belief that p, you've accepted S's testimony in pretty much the same way that you ordinarily accept testimony--all that's happened is that some implicit assumptions have been made explicit. People who don't like that might not like my description of the cumulative case as pure testimonial knowledge. I think that excluding these cases makes us rely too much on blind trust, but your mileage may vary.)

Posted by Matt Weiner at July 12, 2004 03:44 PM

If the subject's justification would involve appeal to the fact of four people all saying the same thing, together with an assertion of the improbability of four people all getting it wrong (which itself requires some belief in the independence of the sources), doesn't that make the knowledge inferential in a way that routine going by testimony is not? It seems to involve the subject being able to give a somewhat detailed argument about the reliability.

Posted by: Anders Weinstein at July 12, 2004 05:41 PM

Maybe. I'm envisioning a case in which the subject might just say, "Well, if they all say it, it can't be wrong," without any elaborate theory of the independence of the informants. I don't think this need be any more inferential than just thinking, "If she says it, she can't be wrong"--both involve a tacit acknowledgment of the extent to which the informant is reliable, but that need not be very far in the psychological foreground.

One of the questions here is exactly what should count as inferential knowledge, and the extent to which it's really important that testimonial knowledge be non-inferential. I think the important distinction is between cases in which you believe what you're told because you think the teller is (to some extent) sincere and authoritative, and cases in which you go through some seriously deviant chain--you believe the speaker is both bluffing and mistaken, for instance. But that's partly because I usually don't focus too closely on the details of belief formation, but rather on the way in which evidence would justify belief.

Posted by: Matt Weiner at July 13, 2004 09:09 AM