# Agreement, Testimony, and Meta-Evidence

Suppose you take S to be a competent inquirer, and you take S to have engaged in competent inquiry concerning p. Suppose as well that these attitudes by you are rational ones. The, if S sincerely tells you that p is true. Some take this to be evidence regarding p, but the most obvious and straightforward account takes it to be evidence for you that there is good evidence for p. That is, the immediate and direct effect of a sincere witness to p is metaevidence regarding p. After all, the point of taking S to be a competent inquirer who has engaged in competent inquiry is that whatever they tell you will be something they have evidence for. So when they tell you that p is true, you obtain evidence that they have evidence for p, which is to say that you get evidence that there is evidence for p.

When such a sincere witness counts as testimony, then metaevidence for p is also evidence for p. That’s why testimony makes one justified in believing what is reported. As noted before, however, having metaevidence for p is no guarantee of having evidence for p. Jamie’s example here is a good refutation:

Suppose that 55% of young guckies are males. And, 60% of young male guckies are aggressive, but only 1% of young female guckies are aggressive, while half of all adult male guckies and half of all adult female guckies are aggressive.

I have Pat, a guckie, in my terrarium. Here is some evidence, which I will call Y: Pat is young. Y is evidence for the following, which I will call YM: Pat is a young male guckie. And YM, in turn, is evidence that Pat is aggressive (which conclusion I call A).

Y is evidence for YM, and YM is evidence for A, but Y is not evidence for A (indeed, Y is evidence against A). So, Y is evidence that there is evidence for A, without being evidence for A. So the meta-evidence principle is false.

I think a similar restriction on the metaevidence principle applies to the word of our peers, both when they agree with us and when they disagree. In fact, the metaevidence principle gives a nice way to characterize the difference betweeen cases of agreement and cases of testimony. When we have a case of testimony, the metaevidence principle is true, assuming that the testimony generates metaevidence; when we have only a case of simple agreement, the metaevidence principle is false. Moreover, this account helps explain the attraction of always taking agreement and disagreement to be epistemically relevant. Since justification by testimony is so pervasive in our rational conception of the world, this pervasiveness leads some to overgeneralize here and take the word of others always to constitute testimony (remember, I’m stipulating that a case of testimony is one where the metaevidence principle is true).

None of this constitutes an argument for my view that simple agreement and disagreement by peers (or superiors) does not count as evidence. But maybe the metaevidence principle helps explain why one would be tempted toward the view I think is mistaken, if you’re convinced that it is mistaken.

#### Agreement, Testimony, and Meta-Evidence — 9 Comments

1. Jon, could you please say a bit more about why the testimony/mere agreement distinction maps into the metaevidence-as-evidence/metaevidence-not-evidence distinction? That is, I would’ve thought the relevant analogy from the guckies case would be cases in which we have additional information that the first-order evidence (for whose existence the metaevidence is evidence of) is in fact misleading. E.g., A finds out that B agrees with her on a question about the ’04 election — but then also finds out that B has lately been reading nothing but frothing-at-the-mouth blogs at a far-out end of the political spectrum. We would have metaevidence of first-order evidence, but also of the problematic nature of that first-order evidence — so, metaevidence without successful first-order evidence.

But this seems to me to be a different issue than testimony vs. mere agreement. Both testimony and mere agreement seem capable of being undercut by further information, as in the case just given and as in the guckie case. So that’s why I’d like to hear more about why you want to invoke the concept of testimony here.

(I might be underestimating the extent to which you are meaning to use “testimony” in a purely stipulative sense here. It had seemed to me that your usage was meant to at least somewhat pick up on our ordinary usage, but that might be a mistake on my part.)

2. Jonathan, oops, I’m assuming that readers of this post have read the previous discussion of agreement here. Some people think that agreement/disagreement by peers/superiors is always evidence for/against what you believe. I don’t think so, and that creates some tension with the obvious point that we learn a lot through testimony. So I dubbed the distinction I’m after–the difference between when the word of a witness counts as evidence and when it doesn’t–the distinction between simple agreement and testimony. So what I was trying to do was, first, to motivate the view that the word of a witness is, in the first instance, metaevidence: it is evidence that evidence exists for what the witness asserts. Then, for such witness to be evidence for the content of the assertion, something more will need to be true. If this “something more” is present, then the metaevidence principle is true, and we have what I’m terming a case of testimony: a case where the word of the witness is not simply metaevidence for p, but evidence for p as well.

My hope is that this distinction will help with a defense of the view I hold by offering an interesting explanation of the stipulated distinction between simple agreement and testimony. Does this make more sense, or do you still find the analogy strained?

3. It seems that Jamie’s case is a counterexample to the claim that evidence of evidence is evidence. But, in the context of the debate, it should just serve as a spur to a reformulation of the principle. Knowing that an epistemic peer disagrees is evidence of something like this: that coming to believe ~p on the basis of shared evidence E is reasonable (better: disagreement by some number of epistemic peers, to allow for the fact of unaccountable mistakes, self-deception, etc). If a putative peer believes that A on the basis of Y. In that case, we have evidence that the believer is not an epistemic peer at all. This line of thought converts disagreement into testimony: it testifies to one way in which it is reasonable to take the evidence.

4. Neil, I like the point that finding out what an epistemic peer does with the same evidence is evidence for what can reasonably be believed on the basis of that evidence. If one is an evidentialist, this point amounts to the view that you acquire evidence about what the evidence shows to be true. So you acquire evidence that the first-order evidence is simply not transparent in its doxastic implications.

I wouldn’t want to push too hard on the possibility of concluding that a person is not a peer because of what they believe in a single case. I expect all of us are guilty of, or at least capable of, some real epistemic howlers, so finding one in another person’s belief system shouldn’t lead us to conclude that s/he is not a peer (or even a superior). It may be very weak evidence for this view, but the weakness of it should not be underestimated.

5. Jon, that does help somewhat. I had followed the previous thread, but my own understanding of the term “testimony” was getting in the way. I think the confusion arose from my thinking of “testimony” as something like “agreement-plus” (i.e., agreement plus your peer’s actually having asserted it to you, or something like that), while you’re using “agreement” as something like “testimony-minus” (i.e., just like testimony but without the juicy epistemic goodness that could really make it evidential).

Not being a huge fan of conceptual analysis, I’m disinclined to fight you for the right to the term “testimony”! But I wonder if maybe this difference of emphasis does have some implications. Namely, you seem inclined to frame the principle in terms of agreement’s needing something more, some sort of enabling condition(s) that would convert it to evidence. I’m inclined to think that it’s more a matter of testimony needing some disabling condition(s) to fail to obtain. Putting it differently: is the principle generally false, except in some special circumstances — or generally true, except in some unfortunate circumstances?

Let me try it my way, as an experiment. I know that there are some good arguments out there (from Peter Klein, inter alia) against an unrestricted transitivity of evidence principle, but has anyone tried to formulate a restricted version? I’m probably reinventing someone’s wheel here, but anyways here goes:

If p is evidence for q, and q is evidence for r, then p is also evidence for r unless:
(i) p by itself would be evidence against r; and/or
(ii) for some evidentially available s, (p&s) is evidence against r.

I think that that works to keep the guckie case as a failure of transitivity. (I’m not sure that it handles cases like Dretske’s zebra case, but I’m not sure that those arise as problems when framed in terms of evidence.)

So peer agreement that p will serve as evidence that p, unless the peer agreement is a case in which (i) the agreed-to proposition is undermined by that very agreement — e.g., Jones and Smith both believe that there is no proposition that they both agree upon — or (ii) in which we have further overriding information at hand — e.g., the politics case in my earlier comment.

6. (I just noticed that actually, (i) is a sub-case of (ii), for any tautologous s. Ah, well.)

7. Jonathan, I think we agree on the conceptual analysis point, and I wasn’t intending to give any sort of analysis of the terms, anyway. I was just inventing labels for the phenomena I’m interested in.

I’m not sure, though, that the transitivity issue is the same as the metaevidence principle issue. I can’t find uniform values for p, q, and r from the metaevidence principle to plug into your transitivity principle. If p is the word of the witness, then p is evidence that there is evidence for some r. So maybe q is the claim that there is evidence for r. But then q is not itself evidence for r, since the claim that there is evidence isn’t identical to, and doesn’t imply that there is some claim t such that t is evidence for r. If q were true, then there’d be such a t; but the case doesn’t require that q is true; only that there’s evidence for it. And, I think the way Jamie’s argument goes is from a denial of transitivity to a denial of the metaevidence principle, and that doesn’t give any promise that we can translate an argument from a true qualified transitivity principle into a true metaevidence principle.

I do agree, however, that the right way to think about the word of a witness to p is in terms of what I’m calling testimony: it ordinarily constitutes evidence for p. I suppose I think as well that this is a contingent fact resulting from the way in which our conception of world depends so centrally on testimonial information. What I’m really after is simply the point that any metaevidence principle that is defensible will have to be qualified in such a way that just going around finding people who agree with you (when you view yourself as competent and your investigation as complete, i.e., the sort that yields the kind of justification necessary for knowledge) isn’t going to enhance your epistemic position.

8. (Jon, I didn’t think you were looking to fight over the ‘right’ usage of the “testimony”, either.)

Here’s the revised version of the metaevidence principle, with my extra condition, letting “Eq” mean “there is some evidence for q”:

If p is evidence for Eq, then p is also evidence for q unless:
(i) p by itself would be evidence against q; and/or
(ii) for some evidentially available s, (p&s) is evidence against q.

I.e, let the disabling condition on the transitivity principle also serve as a disabling principle on the metaevidence principle. I’m convinced by what you say that the metaevidence principle cannot just be an instance of transitivity, because, indeed, it’s not generally the case that Er is itself evidence for r. But Er can, except in bad circumstances, serve as a ‘bridge’ to make the evidence for Er into evidence for r — to make metaevidence into evidence.

Given your stated goal at the end of your comment — which strikes me as very plausible — I think I would take the following line. I wouldn’t attack the revised metaevidence principle per se, and thus I would grant that (except in special circumstances, like those of Klein’s clever car thief) evidence of evidence is evidence simpliciter. So just finding out that a respected epistemic peer says that p will (generally) count as uncovering evidence for p. But: such agreement won’t, even generally, by itself improve your epistemic position w.r.t. p, because to do that it’s not enough for the agreement to be evidence for p. Rather, you’d need also to have evidence that the agreement constitutes additional evidence for p, above and beyond whatever evidence you already have. And it might be that your peer has the same reasons that you already have; you both read this morning’s Times, perhaps. So it’s not more evidence, but the same evidence once again.

Can anything like this worry arises with cases of a peer’s disagreement? Their disagreement is evidence for the existence of evidence that not-q, but suppose you are already aware of some evidence for not-q and have what you take to be good counter-evidence. Then their mere disagreement is not enough to first-order evidence against q. You’d also need to have evidence that their evidence against q is not the same as the evidence against q that you already take yourself to have defeated or overridden.

This problem of unwittingly redundant evidence does not arise for many cases of getting evidence, in that we often can tell that new evidence is, indeed, new evidence. We can tell distinct observations from each other, or two different lines of reasoning. Interestingly (and I didn’t set out to sound like a broken record here!) it can apply in cases of attempted evidential uses of intuition. Namely, it can arise in cases where one’s intuition that p is (tacitly and unbeknownst to the agent) arising from one’s already-held theory about p. One might’ve thought that one has two bits of evidence for p: the theory, and the intuition. But in fact, they are redundant, and one is no better off evidentially with the intuition than without it. (Just one more instance in which intuitions and testimony have deeply similar epistemic properties.)

9. Excellent thoughts, Jonathan! I think you’re exactly right that the issue isn’t so much whether mere agreement gives evidence, but whether it gives additional evidence that would improve one’s epistemic position. I think that’s a better way to put my point, and wouldn’t require undermining some restricted metaevidence principle such as you suggest.

My only residual worry is about testimony itself. One might be tempted to say that you need a reason to think that the evidence base of the testifier/dissenter is different from your own in order for their word to provide enhancing or diminishing evidence. But my worry is that this requires too much of testimony. We could perhaps put the restriction negatively? Maybe that the word of a witness fails to enhance or diminish epistemic standing if you have reason to believe that their evidence base does not include anything more than what you’ve already taken into account. This condition will not be met if you take yourself to be competent and your inquiry to be complete, i.e., to justify the attitude of closure to further inquiry distinctive of the kind of justification necessary for knowledge.