An Interesting Case for Evidentialists

Dale Matthew, a grad student at York, sent me a case that raises an interesting issue for evidentialists. Here’s the case:

It is the consensus of a scientific community that P. In coming to accept P they used all the available resources (theories, observation equipment, etc.) as best as they could. Further, in coming to accept P they were as epistemically responsible as they could be. In other words, they are blameless all around. But then a genius scientist, Insight, arrives on the scene, and through his sheer brilliance is able to discern that instead of P being true, -P is true. Further, Insight is able to discern that the basis on which the scientific community in question came to accept P made P unlikely to be true, highly improbable in fact. Insight relates his, well, insight to the scientific community and the community promptly recants its acceptance of P. Now was the community justified in accepting P before Insight arrived to save the day?

Notice that it is central to the case that the body of evidence is the same for Insight and the rest of the community; what’s different is the assessment error made by the community and the correction by Insight. Since it’s obvious that such cases can occur and that after being corrected by Insight, the community is no longer justified in accepting P, the evidentialist can’t accept the description that the community was justified before hearing from Insight and also hold that justification is a function of total evidence.

I believe this is the issue that Dale hangs his argument on, though he can correct me if I’m wrong. He’s hoping for responses from blog readers, and I’ll hold off commenting on the issue here in the main post to allow the case and the issue to stand on their own.


An Interesting Case for Evidentialists — 19 Comments

  1. “Since itâ??s obvious that such cases can occur and that after being corrected by Insight, the community is no longer justified in accepting P, the evidentialist canâ??t accept the description that the community was justified before hearing from Insight and also hold that justification is a function of total evidence.”

    I think this is a red herring because I doubt that one could ever be in posession of ‘total evidence’. Doesn’t this suggest omniscience? Should we not say that the evidentialist holds that justification is a function of the interpretation of the available evidence given certain interpretive axioms?
    Then what Insight brings to the party is either to demonstrate that the axioms of interpretation were incorrectly applied or, more likely, a difference in the set of axioms.

  2. One thing to notice is that the evidence the scientific community has before and after Insight’s contribution is not the same. After Insight’s contribution to the discussion, Insight’s testimony becomes part of the scientific community’s evidence. So if there is a problem for the evidentialist here, it would have to be that the evidence on the basis of which Insight is justified in believing -P is the same as the evidence on the basis of which the community is justified in believing P.

    But is this evidence the same? It is not clear to me that it is. Presumably after careful consideration of what I will call the first order evidence, E, it seems to the scientific community that P is very likely to be true given E. I think that this seeming should be included along with E in the community’s actual total evidence for P. But of course things do not seem this way to insight. When Insight considers E, it seems to insight that -P is likely to be true given E, not that P is likely to be true given E. In addition, as the case is described, it sounds as if Insight is able to offer an explanation of why things seem this way, and explanation that the scientific community accepts when it hears it. So Insight’s evidence is not the same as the scientific community’s evidence; in addition to E, insight’s evidence includes a different seeming plus various beliefs about why things seem this way to insight.

  3. Mike beat me to the point about seemings but I also think there is a role here for part of what Iâ??ll just loosely call â??rules of evidence.â?? These include both what we take to be (in many cases) synthetic a priori rules about what constitutes evidence and also our subjective theory of evidence consisting in our conditional subjective probabilities (which in most rational people will be effected significantly by our conception of what evidence *is*).

    So the rough-and-ready idea is that if there is a problem here for the evidentialist it could just be that (we) evidentialists need to be more clear about the role of the two evidential factors I mentioned in our very conception of evidentialism.

    My motivation for thinking this is in part as follows. The datum to be explained (away?) by the evidentialist is that the SciCom undergoes change in credence functions while â??evidenceâ?? remains static as does (intuitively) justificatory status. This is supposedly not what the evidentialist predicts. Supposedly since evidentialism makes justification depend only on evidence, a shift in credence without a shift in evidence should lead to a shift in justificatory status. Note, though, that if what happens when Insight relates his insight to SciCom is that either their conception of evidence changes or their subjective conditional probabilities change-as is plausible-then there is no problem for the very broad evidentialist claim (which I think captures the core intuition of most evidentialists) that

    (E) The only factors relevant to assessing whether S is justified in believing p at t are how evidential matters stand with S and p at t.

    even if it causes problems for (and Iâ??m not saying it does) more specific conceptions of evidentialism currently on offer (for example the following ones from Ch. 4 of the recent or forthcoming (Iâ??m working from a MS which Iâ??m not sure has been published yet) Feldman and Connee book on evidentialism:

    (EJ) Doxastic attitude D toward proposition p is epistemically justified for
    S at t if and only if having D toward p fits the evidence S has at t.

    (ES) The epistemic justification of anyoneâ??s doxastic attitude toward any
    proposition at any time strongly supervenes on the evidence that the
    person has at the time.)

    So my suggestion-and it is motivated by factors well beyond the case above-is that (we) evidentialists work on fleshing out a conception of â??evidential mattersâ?? in (E) that takes into account not just that-upon-which-we-conditionalize but also our conditional subjective probabilities themselves (as well as our conceptions of what is evidence for what). This subjective theory of evidence is not usually part of what we intend to refer to by â??the evidence for pâ?? but clearly the set of conditional probabilities of the form P(p/x) are an important part of how evidential matters stand between S and p (see (E)). I should say, finally, that I donâ??t think this suggestion commits me to attributing anything to rational agents much beyond what they clearly do have, namely vague but somewhat fixed dispositions to change confidence in certain ways if they were to learn certain other things. This is one way it could go with Insight. When he explains his insight, they see that certain of their conditional probabilities were mistaken, thus they shift their belief without loss of justification even though â??the evidence for pâ?? remains the same (or so we can assume because in spite of what Mike said-with which I totally agree-someone will come up with a similar case in which appeal to seeming states-my favorite move-will not work). Still, all that has gone into their justificatory staus are evidential matters, so (E) is not threatened (which I’m much more concerned to defend than anything like (EJ) or (ES).

  4. I’m afraid the discussion hasn’t quite focused on the issue I wanted it to focus on. The example is meant to see where people stand with respect to my question at the end — was the community justified in accepting P before Insight arrived to save the day? — and the answer to that question is important because it is a crucial premise in an argument of mine against Stewart Cohen’s famous arugment that evil demon victims are justified. In my view the scientific community was NOT justified before Insight arrived on the scene. It may be that I am the only one that feels that way, but in any case that is the question I wanted to raise.

    Regarding the question of whether the evidence is in fact the same for Insight and the scientific community, it is intended that the evidence be different. It is imagined that Insight sees the inadequacy of the community’s evidence, E, merely through his sheer genius, which he uses to discover a new scientific theory (though that might be hard to imagine, for it is hard to see how even geniuses could avoid taking notice of scientific discoveries that precede them), which in turn shows that P is highly improbable on E.

    As regards subjective probablities, I see them as epistemological non-starters (for reasons I will not get into). In fact, my whole project is to move epistemology towards a more objective understanding of justification.

    In closing, I’ll pose this question: If the scientific community was justified in believing P before Insight arrival, what is the source of this justification (what gives the belief its positive epistemic status)?

  5. Dale, if one is an evidentialist, then the answer to your last question is “the evidence.” And the fact that there is some notion of probability on which Insight discovers that P is improbable on the evidence only shows that that notion of probability is not decisive in determining justificatory status. Given your question, I think you may think the example has more probative value than the responses above take it to have. Or is there more to your argument than the specific example?

    By the way, the new evil demon problem appeared earlier than Cohen’s argument. Foley talked about it earlier, and, having learned it from Foley, I talked about it earlier as well, in a 1983 publication. I think the original source may be Sosa, but I’m not sure (maybe in “The Raft and the Pyramid”?).

  6. Well I guess Mike and I did misinterpret the import you intended (I liked Mikeâ??s response and I think we interpreted it pretty similarly), but I must say it worked well to illustrate some important ideas. In my post I took it as a given that SciCom was justified in their belief that p as I saw no reason to suppose otherwise. Itâ??s a particular instance of a very common phenomenon: justified false beliefs. Since this is a common phenomenon (says a wide consensus) youâ??ll have to tell a story about why this case would be any different. Merely having been wrong isnâ??t adequate evidence that they werenâ??t justified. Nor is it adequate that the evidence objectively supported the negation, at least not for the subjectivist about justification. So if you want to mitigate against subjectivist justification, youâ??ll either have to take a more frontal assault or expand the story. The subjectivist has any number of answers as to why the SciCom was justified before. My preferred version is that given their subjective theory of evidence, their personal probabilities were high for p on the set of evidence they had. A credulist could say that they were justified because p seemed true to them and they had no justification defeaters. A deontologist could say that they violated no epistemic norms in believing p. So it seems to me there are a lot of good options for an answer to your question, itâ??s really just a matter of which one is the best.

  7. My idea was that the example would elicit agreement so that you’ll have not problem accepting the first premise of my argument against evil demon arguments (by the way I saw the example in Foley’s The Theory of Epistemic Rationality and so I wouldn’t be surprised if they appeared in his articles leading up to the book, but Cohen’s paper seem to be the one most cited in evil demon world discussions):

    Stewart Cohen says the following:
    â??Certainly[,] if we were to know that our cognitive processes are unreliable[,] then the beliefs they generate would not be justifiedâ?? (â??Justification and Truth,â?? p. 281).

    I take the idea expressed here â?? that beliefs formed in a manner or on bases that we know make them unlikely to be true are not unjustified â?? to be obviously true and unlikely to generate much dispute. Thinking of this idea led me to the following idea, which is the first premise in my argument that to be justified a belief needs to be formed in a manner or on a basis that makes it likely to be true and that as a result the beliefs of evil demon victims are not justified:
    1) Beliefs that we come to know are formed in a manner or on bases that make them unlikely to be true were not previously justified before we came to know that they are unlikely to be true.
    2) If beliefs that we come to know are formed in a manner or on bases that make them unlikely to be true were not previously justified before we came to know that they are unlikely to be true, then at least one reason why they are unjustified is that they are unlikely to be true.
    3) Thus beliefs formed in a manner or on bases that make them unlikely to be true are not justified.
    4) Thus beliefs need to be formed in a manner or on bases that make them likely to be true in order to be justified.
    5) The beliefs of evil demon victims are not formed in a manner or on bases that make them likely to be true.
    6) Therefore, the beliefs of evil demons victims are not justified.

    I was hoping that by the example I wouldn’t have to have an extended dicussion of 1). But since that obviously hasn’t worked out I’ll say something briefly about 4, which what I really want to defend. It just seems to me that subjective accounts of justification are just variations on the idea — call it “B” — that “P is justified for S if S believes that S’s evidence for P, E, is good evidence for P” and that is just too subjective in my view. That is, unless you are objectivist about “good evidence” — whatever it is that gives a belief a positive epistemic status — and believe that the justification is a matter of beliefs being being connected in some objective way to truth (as Goldman’s reliabilism does, though my externalism is quite weak) you have to accept B — but B is clearly untenable. Now I guess the issue now is: refelecting on B, do subjectivists have to have to acknowledge that their accounts are variations on B?

  8. Dale, I think you’ll find externalists inclined toward (1) and internalists inclined against it, and the new evil demon problem is normally thought of as an argument that is sympathetic to internalism. So an argument using it to undermine the new evil demon argument won’t be very persuasive.

    I’m not sure what the boundaries are on a theory being a variation on a theme, but it is important to note that some internalists are subjectivists, but not that many. That raises the question of the truth connection for internalists who are not subjectivists. As I’ve posted here before, internalists can adopt the Chisholmian idea that, in rough terms, the justifiers of p are also justifiers that believing p is the appropriate thing to do for a purely intellectual being, or that p is epistemically justified, or that believing p is appropriate when your goal is to believe the truth and avoid error. This issue is important for internalists, I think, because externalists have a simple account (albeit encountering difficulty with the new evil demon problem), and some account of how epistemic justification is tied to truth is required of any account of that notion. Internalists have not been forthcoming in this regard, however.

    Here’s a general line of argument against 4, though I’m going to put it in somewhat of a rough and ready form. Objective probability relations are just more facts that a person may or may not be aware of. In general, justification is perspectival, varying according to the facts and information within the purview of the individual in question. So if you know the objective probability facts, that will tend to undermine justification, and if you don’t, they won’t have any more relevance to justificatory status than other facts about which you have no information.

  9. Dale, if 1. involves a sufficiently coherent notion of ‘justified’ to make sense, it would seem that it (1) is either a) false or b) true, which, given your initial example, suggests that no belief is justified or in principle justifiable.

    I have a question concerning adherence to (4) or something like it: once we accept a suitably ‘objective’ account of ‘good evidence for P,’ something I suspect will eventually bottom out in ‘P is true,’ what philosophical work can we expect from the notion of justification? Won’t we simply be left with reports of true and false beliefs? What further thing would we be saying by the claim that someone’s belief is not only true but ‘justified’?

  10. I think that given the connection that there must be between justification and truth — something like what reliabilists say, or what I said above — we are going to have to give up the perspectivism. I did not start out this way when I started to think about epistemology, but I think that is the only way to take seriously the goal of truth. I think there are internalist constraints on our idea of a person being justified, but I think justification goes beyond our “own state of mind,” to use Chisholm’s way of putting it. (But Jon, I don’t get what you mean when you say “So if you know the objective probability facts, that will tend to undermine justification.”)

    If “P is justified” were to bottom out in “P is true” that would indeed be a serious flaw in any account of justification, but I don’t see why objectivism about justification should have to come out that way. Why should objectivists have to go beyond, say, what is objectively probable? Still, if what I am saying about the implications of taking the truth goal seriously is true, the work of espistemologists would have to be signifantly qualified. The question of whether P is justified would be dependent on facts we may not know, and so we’ll just have to be content with saying something like “S is justified in believing P if …”, where “…” is some fact of the matter that we might have no knowledge about.

    But before we get to that point I’ll just have to work on establishing 4), for its clear to me now that only people who accept it — who are, as jon says, most likely externalists — are likely to accept 1).

  11. It would be unfortunate for externalists if their view required giving up perspectivalism. That would be about as decisive a refutation of the view as one could ask for in epistemology. In fact, extant versions of externalism are all crafted to preserve perspectivalism, so if defending (4) requires rejecting them as well, then you’ve got a position with no natural allies. Of course, that’s not an argument against (4).

    I also think it is too much of a leap to think that endorsing (4) is the only way to take seriously the goal of getting to the truth and avoiding error. There are quite a number of options for taking that goal seriously that reject (4).

    The remark you ask about is nothing more than an affirmation of perspectivalism: if you discover that a proposition you believe has a low objective probability, that tends to undermine whatever justification you have for it.

  12. Dale,

    If you’re surveying people’s intuitions, my sense is that the scientific community was justified in their belief. Actually, that seems so right to me that when I first read the example, I thought you were giving it in order to refute the idea that justification requires objective probability of truth.

    Perhaps I can pump some intuitions a little. Imagine you are a scientist around before Insight comes along. The accepted theory is P. What should you believe? You have 3 options: believe that P, deny that P, and withhold judgement. To you, like everyone else around, it seems that the evidence clearly supports P. What is most rational for you to do?

    Surely it wouldn’t be rational for you to deny P. If you denied P, what rationale could you possibly provide? You’re not in a position to provide whatever rationale Insight is going to provide because, by hypothesis, you don’t have the insight required to see that. So it seems that you’d just be irrational if you denied P.

    Withholding P might seem less irrational, but still: what rationale could you provide for why you are withholding rather than believing P? Well, you could always say that the evidence isn’t 100% conclusive, but that is always true, and it’s not the case that it’s always rational to withhold with respect to scientific theories. Could you say that the evidence doesn’t make P sufficiently likely? No, because it seems to you that it does. By the construction of the case, you’re not in a position to actually give any rationale for either withholding or denying P that would not apply to cases in which belief is clearly justified. So I think you’re justified in believing P; indeed, I think it would be irrational for you not to believe it.

    In answer to your question as to what justifies P, I believe the following epistemological principle: “If it seems to S that P, then S is thereby at least prima facie justified in believing that P.” (I call this “phenomenal conservatism.”) Note that the principle is not that “If S believes that P, S is thereby prima facie justified in believing P,” which, in my view, would be a bad principle. I think, then, that the scientists are justified in believing the following:

    The evidence that we have makes it likely that P.

    because that seems to them to be the case, and they have no grounds for doubting it (again, by construction of the example: if either that doesn’t really seem to them to be so, or it does but they have serious grounds for doubting it, then I take it the example loses its interest). Now, the above proposition, itself, supports P (even if the evidence referred to does not).

  13. Hi Dale,

    The case is interesting, but it seems to me to be a problem that it presupposes (i) that in coming to believe P and the basis on a body of evidence the scientific community were as epistemically responsible as they could be AND (ii) that Insight did ‘better’ on the basis on the same body of evidence.

    Intuitively (i) and (ii) are in conflict. One would think that if the scientists did as well as they could given the evidence they got, then one could not do better given the same evidence. Indeed, it appears that Insight simply reveals that the scientists assesment of the evidence is not ‘as responsible as it could be’.

    Of course you could make it the case that the scientists were very epistemically responsible, but less-than-perfectly responsible. Perhaps, in such case the evidentialist might reply that the scientists were indeed justified , but that their justification was defeated by the more insightful take on the evidence?

    Don’t know whether this is helpful. I found it hard to evaluate the case because it is put in the abstract. To my mind a concrete case spelling out what the evidence is and how the scientists came to belief that p on the basis of it (and how Insight did better) would be much more powerful.


  14. The idea I had in formulating the example is to make the best possible case for the other side, and to preclude certain potential objections (for example, that what makes the scientists unjustified is that they were epistemically irresponsible, though I think this way of thinking about justification is mistaken). Unfortunately, it does not seem that doing so worked in my favor.

    To Mikkel, as I noted in message 5 above, I would deny that the community and Insight have the same evidence. Insight’s evidence, E+1, includes the community’s evidence, E, plus ‘something extra’ (1). And the way that Insight’s something extra doesn’t run afoul of the no-epsitemic-irresponsibility condition regarding the community, is that he comes in possession of this extra something in such a way — through the application of his extraordinary intellect to the problem at hand — that a failure to do so does not signal any epistemic failures on the part of the scientists in the community.

    It seems to me that “phenomenal conservativism” — according to which â??If it seems to S that P, then S is thereby at least prima facie justified in believing that Pâ?? — is just the wrong way to think about justification. By saying this I am taking full account of the “prima facie” aspect of what you say. That is, I do not think that it seeming to S that P even prima facie justifies S in believing that P. The reason is that in principle any proposition can “seem” to S, even the most bizarre and unlikely of propositions. In other words, it seeming that P to someone doesn’t have any connection to the goal of truth. It could seem that P to S for reasons that make P the most unlikely of propositions, and surely in such cases we don not want to say that S is justified.

    I would admit, however, that in the circumstances of the scientists, believing P might be the most rational think for them to do. But I do not thereby admit that they are justified in doing so. This is because justification requires something that is external to the observer, namely, an appropriate connection to truth.

    About perspectivalism, what I meant above was I reject the the view that justification is WHOLLY a matter of what’s goes on in a first-person perspective, not that nothing in a person’s perspective is relevant regarding justification (as should be clear given my acknowledgment earlier that there are internal constraints on justification).

  15. Dale, I realize that you didn’t mean to deny perspectivalism completely. But unless you have a completely theory-driven response, you’ll have problems sustaining your approach here. The probabilities you treasure are contingent, and like most all contingent claims, they are not knowable a priori. So if other contingent factors like this that can affect justification once one is aware of them do not affect justification when one is not aware of them (and to deny this point is to deny perspectivalism wholesale), what make the contingent facts about probabilities or likelihoods different? Why is their mere existence sufficient to undermine justification when other factors with the same contingent status are not?

  16. Dale,
    I don’t think that the most unlikely of propositions can seem to S to be true while S has no grounds for doubting them–but that’s just because I accept an epistemic interpretation of probability. You’d be right about physical probability. However, I think it is still false that:

    it seeming that P to someone doesnâ??t have any connection to the goal of truth

    Its seeming to S that P is connected to the goal of truth in the following way: If it seems to S that P, then (other things being equal) it would naturally seem to S that the best way of pursuing his goal of believing more truths and avoiding error is to believe that P. Therefore, believing P would constitute doing his best to further that goal; any rational person who was trying to further that goal would do that (other things being equal).

  17. Dale,

    I’m with you that a non-accidental ‘truth-connection’ is required for any kind of epistemic warrant. But I’m with Mike Huemer as far as the case goes. If the scientists are indeed ‘as epistemically responsible as they could’ in believing P on the basis of E, then they are being as responsible as they could in holding a true belief. I take it that this is what epistemic responsibility amounts to. So if one hold a belief in accordance with truth-conducive norms, one’s belief is indeed ‘connected to truth’ although fallibly so.

    Likewise, insofar as the rationality which you grant the community is a kind of epistemic rationality (a kind of rationality having to do with holding true beliefs), then it seems to suffice for (a kind of) non-factive justification?

    I get from your qualification in Remark 5 that E+1 isn’t simply empirical evidence undermining that E justifies P (such as testimony to the effect that the zebra looking animals are in fact disquised mules). But I am uncertain what E+ amounts to. If E+ is a novel and superior theory according to which E justifies non-P, then it does seem that, the community is less-than-perfectly responsible (because they have failed to consider such a theory viz-a-viz E and P). Their inability to asses E in the light of Insight’s theory may not make them epistemically irresponsible (this is required for the case). But all the evidentialist needs is that they are ‘not as responsible as they could be’. And non-irresponsibility does not entail perfect responsibility. This is because epistemic responsibilty and the kinds of justification associated with it plausibly come in degrees.

    Again, insofar as the case is supposed to appeal to INTUITIONS, it appears to be a little underdescribed. In particular with respect to what, precisely, the difference between Insight and the community amounts to.

    Finally, I should add that I’m a pluralist about warrant. So what I believe is merely that there is a sense in which the community is epistemically justified.

  18. Dale says:

    â??I think that given the connection that there must be between justification and truth â?? something like what reliabilists say, or what I said above â?? we are going to have to give up the perspectivism.â??

    The statement â??p is trueâ?? is, as Ayer pointed out, logically equivalent to â??pâ??, and Ayer wrote to the effect that reliabilism is precisely the criterion we should use to verify p, but he also pointed out, well in advance of his critics, that there is a perspectivist element to verification â?? â??A man can always sustain his convictions in the face of apparently hostile evidence if he is prepared to make the necessary ad hoc assumptions.â??

    Jonathan says

    â??So if other contingent factors like this that can affect justification once one is aware of them do not affect justification when one is not aware of them (and to deny this point is to deny perspectivalism wholesale), what make the contingent facts about probabilities or likelihoods different? Why is their mere existence sufficient to undermine justification when other factors with the same contingent status are not?â??

    So there is some debate as to whether belief is a wholly subjective phenomenon? Assuming that it is, then how can any contingent fact of which the subject is unaware have any bearing upon her justification of her belief? I think that a belief is justified iff the subject could give a rational explanation of why they hold to it in terms of their evidence and their assumptions or, in other words, if they could show that it is consistent with their evidence and their assumptions. Dale seems to think that the justification for a belief should also depend upon the objective likelyhood of the beliefâ??s being true – I hope that I am not misrepresenting him. We both, it seems to me, insist on the centrality of evidence in justifying belief, but differ in our beliefs about the appropriate domain of evidence to be considered. Oneâ??s view of whether the beliefs of the evil demon victim are justified would need to take account of this additional requirement.

    I think that the original scenario gives some ideas as to which is the more useful view. The scientific community which believed p also believed that p was objectively the most likely. Following Insightâ??s insight they changed their collective mind and considered ~p objectively more likely. One can readily imagine Son of Insight coming along with a new synthesis of the evidence and convincing them that p again. What then is objective likelyhood? If I know that the holder of a given belief is an evil demon victim, should I say that his beliefs are not justified because he does not have access to the objective truth, or might it be that I too am an evil demon victim?

Leave a Reply

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