What Defeats Testimony?

In the previous post Jon writes:

I’m thinking of the following kinds of cases: empirical disputes, political disputes, moral disputes, etc. There are two, er, bizarre views here: the first is that unanimity of testimony is required for rational acceptance of the content of the testimony, and the other is the position… that no amount of disagreement matters.

How about this viewpoint? Testimony creates a presumption in its favor that requires defeating information to block rational acceptance, and the defeating information must itself be something more than that there is contrary testimony. What else? I can’t say, but I can list some things: some sketch of an explanation of why the testifier is mistaken or unreliable; the existence of contrary testimony plus a claim to the effect that there are no rational grounds on which to prefer one piece of testimony to another; etc. Central to the position is that testimony is a generative source of justification (or rationality or (non-Plantingian) warrant). Many here have worked on testimony much more than I, so I ask: does this position withstand scrutiny (or better: is there is precisified version of it that withstands scrutiny)?

My view is that a precisified version will withstand scrutiny, but that we need to distinguish a whole bunch of distinctions first. The question seems to me very sensitive to what exactly we mean by justification and rational acceptance. Maybe the best way to explain is to start with the Thesis

(T1) Testimony creates a presumption in its favor that requires defeating information to block rational acceptance, and the defeating information must itself be something more than that there is contrary testimony

and then precisify in stages.

First stage: What do we mean by ‘rational acceptance’? If we mean that it is rational for the hearer to act on the assumption that what she is told is true, then I don’t think T1 can hold, because there will be cases in which it isn’t rational to act on the basis of unsupported testimony. Consider Fantl and McGrath’s example in their Phil Review paper: If it is absolutely vital that you make it to Foxboro soon, the unsupported testimony of a stranger won’t be enough to make it rational to act on the assumption that this train stops in Foxboro; you should double-check.

But that doesn’t mean that the stranger’s unsupported testimony is epistemically worthless; it just doesn’t have enough worth to make acceptance rational given the high stakes involved. So in this case (perhaps) the testimony provides some evidential support. First precisification: talk about evidential support rather than ‘rational acceptance:

(T2) Testimony that p provides some evidential support for p unless there is some defeating information, and the defeating information must itself be something more than that there is contrary testimony

[What I’m invoking here is a bit like the distinction between pro tanto and on balance justification that Peter Graham cites in his paper Liberal Fundamentalism and Its Rivals, forthcoming in the Lackey and Sosa volume on the Epistemology of Testimony. And maybe it will console Jon a bit about the pragmatic encroachment on epistemology–even if pragmatics encroaches on rational acceptance, evidential support might remain free of pragmatic encroachment.]

The second question is what we mean by evidential support. If evidential support is viewed externalistically, then (T2) will fail. Take a view on which whether something provides evidential support is a matter of whether it brings the believer closer to knowledge (with lots of qualifications about Gettier cases, true belief, etc.) On this view the testimony of a stranger will not provide evidential support if the stranger happens to be a habitual liar, or incompetent about the subject matter on which he is testifying, even if he was right in this case. The testimony does not bring us closer to knowledge. On such a view perhaps testimony at best passes a warrant along from speaker to hearer–though the hearer might be blameless for believing the speaker, if the speaker had no evidential support for the testimony the hearer doesn’t have any either.

But we can view evidential support internalistically. Then the question is, given the fact that I have just been told that p (and whatever other information I may have on the subject), what degree of credence should I give to p? Here it is plausible that testimony from someone who is in fact incompetent or a chronic liar still provides evidential support, so long as you don’t know of your informant’s flaws. So precisification 2 is to specify that we’re taking things internalistically:

(T2) Testimony that p provides some evidential support for p (viewed internalistically) unless there is some defeating information, and the defeating information must itself be something more than that there is contrary testimony

Now the question is what kind of information can defeat the testimony? The fact that there is some contrary testimony won’t in general defeat the testimony, or all testimony will be defeated. On the other hand, two equal opposing testimonies must defeat each other if there are no grounds for breaking the symmetry between them. In at least one comment below Jon suggested that one difference might be that one piece of testimony has been received and the other not. That is, if I’ve been told that p (and have no reason to doubt this testimony) and know that someone else is saying that not-p but haven’t heard it myself, in general I’ll have evidential support for p. As I read Jon, this is because I haven’t been able to monitor the testimony that not-p to see if it betrays any signs of untrustworthiness.

I think that there will be cases in which this won’t hold. Take this case:

JUROR. I am a juror at a trial concerning some arcane technical issue on which I have no expertise. The plaintiff calls an expert witness who testifies that p. Though I cannot make head or tail of the witness’s testimony, the witness seems as calm and credible as I would expect an expert witness to be. However, I know that the next day the defendant will call a different expert witness, who will testify that not-p.

Evaluating the juror’s beliefs in the time between the testimony of the two witnesses, I don’t think the juror has any evidence in favor of p as opposed to not-p. The first witness was no more and no less credible than the juror expects a witness to be; so the juror should expect the second witness to be exactly as credible. Given that the juror expects to find himself in a situation in which the symmetry between the testimonies is not broken, and so there is no evidence in favor of p or not-p, he (it seems to me) has no evidence in favor of p or not-p now.

The case would be very different if the first witness were an expert who had been selected at random and the second witness were called by the defendant. In that case, the juror would have the following information:
(1) A randomly selected expert says that p.
(2) An expert selected by the defense will say that not-p.
And (2) more or less amounts to: There is some expert who will say that not-p. For the defense is naturally going to pick whatever expert will say that not-p. So (1) and (2) together provide more evidence that p than that not-p; assuming, always, that there isn’t some other reason to believe the second expert rather than the first.

In the original Juror case we have one of the defeaters Jon cites: The existence of contrary testimony plus a claim to the effect that there’s no reason to prefer one piece of testimony over the other. In the second case we just have the existence of contrary testimony. And in the second case the contrary testimony doesn’t defeat the original justification.

Furthermore, the existence of contrary testimony requires at least justification for the claim that there’s no reason to prefer one piece of testimony over the other. On the internalist point of view, if we’re not justified in believing that claim, then we will be justified in preferring one piece of testimony, and so that testimony will yield undefeated epistemic strength. Of course other things besides contrary testimony can defeat the original testimony, such as the explanation why the teller is mistaken or unreliable (as Jon mentioned).

I’ve already alluded to an interesting question here: Is actually received testimony in general to be preferred over testimony that hasn’t been received? If we’re good at detecting insincere or unreliable testimony, then the very fact that we haven’t detected this in the testimony we have received gives us reason to prefer it over the testimony we haven’t received; we don’t know whether, if we were to hear the other testimony, we would detect something wrong with it. What goes on in the JUROR case is that the hearer isn’t good at detecting unreliable testimony in these cases, and so has no reason to prefer the already received testimony to the tesitmony that hasn’t been received yet. I myself am a bit skeptical about how good we are at monitoring the reliability of testimony, so I necessarily think that JUROR-like cases might be more common than we’d think.

But I do think that, precisified as I’ve specified, Jon’s thesis is plausible.

I should say, however, that I’ve simply assumed and not argued for the idea that testimony does have some default evidential strength. Lots of people working on testimony criticize that. In It Takes Two To Tango (in Lackey and Sosa) Jennifer Lackey criticizes this idea. I think that the criticisms depend in part on her externalist notion of justification, but this post has already gone on far too long….


What Defeats Testimony? — 3 Comments

  1. Matt, the Juror case is very interesting, and persuasive. I have one minor suggestion, though. I think when you hear the first expert, it’s not so much a question of whether you are good at detecting unreliable testimony as it is a question of our self-conception as truth detectors in such a situation. I think you granted something like this when you moved to evidence internalistically conceived in your second precisification of T1. Even given that point, though, the juror case is very interesting, since it’s just the sort of case where we aren’t confident of ourselves as truth-detectors anymore. Of course, it would me happiest of all if such subjective defeaters explained all that needed to be explained, but that’s not something I have a good reason to believe at this point…

  2. Jon, that’s an interesting question. I’m trying to be neutral among various internalistic conceptions of justification here, and I think on some of them the juror case depends on our self-conception as truth-detectors, and on others it depends on whether we have any evidence that we’re good truth-detectors. On the second kind of case what matters is, more or less, whether a reasonable person who’d experienced what you had would think she was good at detecting unreliable testimony in this area. (Er, modulo the counterfactual fallacy I just committed.)

    I take it that on your view what matters is what the juror thinks about the first witness. If I think, “Well, I can tell that this person knows what he’s talking about,” then I am more justified in believing p then ~p, at least until the second witness shows up and I think “Gosh, this person also knows what he’s talking about”? For it wouldn’t make sense to have those thoughts and to completely suspend judgment on p.

    On that view, it is important that the Juror actually can’t make head or tail of the expert testimony, and so doesn’t see himself as a good truth-detector. But I also need it to be the case that the Juror’s lack of confidence doesn’t defeat the testimony all by itself. And I think that’s at least plausible–if the expert witness produces a lot of mumbo-jumbo, I can say, “Well, I don’t follow the argument, but the fact that this expert thinks so is enough to give me some evidence for what she’s saying.” That would be defeated by the existence of the next witness.

    I think the case can be transposed pretty well to other sorts of views, though–the Juror hasn’t actually heard anything that would enable him to tell whether this witness is more reliable than the next, so he doesn’t actually have any justification for favoring the witness over the next before hearing the next. All we need is some sort of consistently applied internalism.

  3. Matt, I like the sensitivity to problems with counterfactual accounts!

    The one thing troubling to me about the juror case is that it forces us to adopt a more reflective mode than we typically do in cases of testimony. That’s especially true when faced with expert testimony knowing there’ll be a contrary witness tomorrow. In the ordinary, unreflective case, I think the default setting is that we trust ourselves as truth-detectors and thereby trust what others tell us in the absence of clues of inaccuracy. As soon as we reflect, things get more troubling, since most of us are aware of how easily deceived we are. So we get the not surprising result that philosophers make lousy jurors, and epistemologists are probably the worst of all–it’s akin to Lewis’s hyperbole that once we reflect on whether we know, we lose all of it!

Leave a Reply

Your email address will not be published. Required fields are marked *