There’s no such thing as defeat

Or, more cautiously (sometimes I’m cautious), I find defeat talk misleading.  And here’s why.

Let q be the true proposition that S gives testimony with content p.  This is a reason for me, at t1, the time of my hearing the testimony, to accept p, since I’m aware that S is generally reliable.  But then I learn, at t2, d: that in spite of his general reliability, S is known to be unreliable when it comes to p-type matters.

In defeat language, d defeats the support q gives to p. But what can this mean?  What has been defeated?  Certainly not the connection between my original evidence (captured in E1 below) and p.  Consider the following two sets of evidence.

E1: S said “p” & S is generally reliable.

E2: E1 & S is unreliable concerning p-type questions.

E1 characterizes my evidence at t1, and E2 characterizes my evidence at t2.  At t1 my total relevant evidence supported p.  At t2 it did not.  But what is it that has been defeated?  Not my justification at t1!  Justification is a synchronic matter.  My transition from being justified in believing p to not being so is a diachronic matter.

And even at t2 E1 supports p.  Its support for p is an objective relation it bears at all times if at any time.  That relation doesn’t fail to bear just because it’s part of a larger body of evidence such that that larger body of evidence doesn’t bear that relation.

In short: If there is defeat, then there is something that is defeated.  But there is no viable candidate for anything that has been defeated.

I followed Chisholm in defining defeat over a support relation (3rd ed, 55).  But I don’t see that it’s any better following Bergmann in defining defeat over target beliefs or propositions.  My justification at t1 was never defeated.  I just got new evidence and on that evidence I was no longer justified.  Nothing is defeated.

We sometimes talk about one item of information “screening off” another, but I think that’s more helpful when thinking about causation.  Perhaps defeat talk is innocuous, but I find it misleading and sometimes (today) it irks me.


There’s no such thing as defeat — 19 Comments

  1. Trent,

    Interesting post. Like you, I think there’s some strange stuff in the defeat literature, but let me take a stab at defending a view in the area of the view you discuss. (I’m not married to any of this by the way, I just don’t want to read what I have waiting for me when this comment is done.)

    You agree that at t1, I can say, “Given what I have to rely on, I have good support for p” and at t2 I can say, “Given what I have to rely on, I do not have good support for p”.

    I agree that, “If there is defeat, then there is something that is defeated”. So, what’s defeated? The support provided by what I have to rely on. It’s true that the things that constitute the thing I have to rely on @ t1 and t2 differ, but why not say that the relation between the particular bits of evidence and what I have to rely on is sort of like the relation between a statue and its clay. You can add bits, you can take bits away, but the statue and what I have to rely on can persist from one time to the next and what gets defeated is the support provided by the thing that persists. It’s sort of like if you take away bits from the statue that is holding up some books. By taking away some of its bits, you undermine the support the statue provides to the books.

    In both cases, we have one thing at different times (i.e., the statue, what we have to rely on), the one thing at different times provides different levels of support, and it does so because the elements that constitute these things have changed over time.

    So, you’re right that some of the relations between constituents hold both at t1 and t2, but that’s consistent with saying that the whole that has these constituents as parts stand in different kinds of relations to the target proposition or propositional attitude. Again, in the statue case, if you subtract a bit of matter or add the wrong sort of matter, the parts that supported the books might continue to do so over time even if the way the whole supports over time changes.

  2. Hey Trent,

    What gets defeated is prima facie justification. This may be clearest with respect to doxastic justification, but it also holds for propositional justification. Suppose I base my belief that P on E1. My belief that p is ultima facie doxastically justified. Later, I still base that belief on E1 even though I now have E2. Basing my belief that p on E1 still provides my belief with prima facie doxastic justification, but not ultima facie doxastic justification.

    Note: in saying that prima facie J is what gets defeated, I don’t mean that when it is defeated you don’t have it. I mean only that the prima facie J is prevented from being ultima facie J.

  3. Hi Trent…One unexamined assumption you might consider is that S being unreliable concerning p-type questions does not count against S’s general reliability. This is one assumption of your idea. However, what if the p-type question is relevant to S’s general reliability? For example, what if S is unreliable with regard to questions of causation? When questioned about who was at fault in an observed accident S reverses the order of causation so that the innocent becomes the accused. So, we must assume p-type questions are trivial in that they don’t impact S’s standing as generally reliable. However, if p-type questions are trivial with regard to impact on the overall ratio of truth to falsehood, then at t2 the total relevant evidence still supports p. This is because S’s unreliability about p-type questions is weak evidence against p. Arguing counter to this requires assuming unreliability about p-type questions does impugn general reliability. But if this is the case, the idea falls apart.

  4. Chris T.

    re: “Prima facie” justification.

    Like Kagan, I find this talk misleading and confusing. Kagan, The Limits of Morality, 17. Oh nice, the quote is on Wikipedia: “”

    I have no idea what prima facie justification is. And E1 remains–eternally–a pro tanto reason for p, even at t2.

  5. Thanks Trent. I thought I created a dilemma for the idea. Either way you stipulate there’s a problem:

    If p-type questions = trivial (regarding general reliability), then at t2 the total evidence still supports p [So, E2 doesn’t work as you suppose].


    If p-type questions = not-trivial, then it’s not the case that S is generally reliable [So, E1 doesn’t work as you suppose].

    So, the idea fails to be useful when looking at questions of defeat. Or, perhaps I’m missing something…

  6. Trent,

    This remark, “And E1 remains–eternally–a pro tanto reason for p, even at t2” seems to miss my note in my first comment. (It would have been better if ‘pro tanto justification’ were term used, but ‘prima facie justification’ is what stuck.

    In any event, if you are not sure what prima facie/pro tanto justification is, here is a suggestion. (I presuppose that you are familiar with Lewis’ 1970 J Phil paper on how to define theoretical terms.) Suppose that we have some independent understanding of what ‘(ultima facie) justification’ and ‘prevents’ are. They will be O-terms in Lewis’ parlance. We can then use those O-terms to define a theoretical role for ‘prima facie justification’ and ‘defeater’. The defeater role can be defined by this sentence: “X is a defeater just in case it is something that prevents a prima facie justification from constituting an ultima facie justification.” The prima facie justification role can be defined by this sentence: “Prima facie justification is ultima facie justification just in case it isn’t prevented from being so by a defeater.”

    (Yes, ‘prima facie justification’ and ‘defeater’ would be partly defined in terms of each other, but this is allowed in the Lewisian way of defining theoretical terms.)

  7. Oh, and some additional benefits of my way of thinking about defeat:

    1. It applies equally well to propositional and doxastic justification, where Bergmann’s view seems most natural when the subject actually has the belief that P.

    2. It applies equally well to internalist and externalist views of justification. Some accounts of defeaters say that they defeat reasons (I believe Pollock’s account is in this category). But if you are an externalist, you might think that a belief can be justified without reasons. In such a case, there would be nothing to defeat on the reasons view. That seems like the wrong result.

    In short, my view gives a clear interpretation to ‘defeater’ and ‘prima facie justification’, and it avoids the problems with all the views that say what gets defeated is support, reasons, and belief.

  8. Chris C,

    I don’t think the dilemma works. Just think statistically–I know I always say that.

    n/m of my statements can be true in the total set of statements I make. But there will be subsets of my total statements corresponding to fairly natural topics where n/m-e of them are true.

    I have a friend, for example, who’s generally quite reliable. So simply his say so counts in favor of a proposition. But he’s totally unreliable when it comes to when and where meetings are.

    So I’m not saying anything fancy, just stuff we’re familiar with in everyday life.

  9. Chris T.,

    First, I remembered that Tom Senor has a paper on the prima facie distinction in epistemology. I hope to re-read it sometime soon.

    Now your suggestion is interesting and creative. You people make blogging worth it!

    However…I just don’t know what I think about this application and the circularity in this case. I suppose I’m inclined to say something like “If THIS is all we can say, then why bother saying it? I’ll stick with my evidence-on-total-probability” talk.” The concept of a defeater was supposed to be super-enlightening. This is not so much, as far as I can tell right now. I’ll have to think more about it though.

    Re: the follow up post.

    1. That would indeed be an advantage over Bergmann’s view, in my view.

    2. I *suppose* this is *some* kind of advantage. But I’m unmoved by consistency with obviously false views. 🙂

    You state the thesis well in your last line, now go build a paper around it! (Seriously.)

  10. Hey Trent,

    You say, “I suppose I’m inclined to say something like “If THIS is all we can say, then why bother saying it? I’ll stick with my evidence-on-total-probability” talk.” The concept of a defeater was supposed to be super-enlightening. This is not so much, as far as I can tell right now.”

    The concept of defeater is really only going to be useful if you talk about small bits of one’s total evidence providing you with justification. One might want to allow some small part of one’s evidence to provide justification, because I doubt that I ever base my beliefs on my *total* evidence. If you don’t want to talk about doxastic justification, then you probably will have little use for the concept.

  11. Christ T.

    I was indeed talking about normative defeat. In fact, it seems to be a thoroughly normative concept, so I’m not sure what it would mean to apply it outside the context of propositional justification. Some people talk about knowledge defeaters, but I don’t remember what they supposedly mean by that. Propositional justification is the core notion, at any rate.

  12. Trent,

    Based on your response, I’m not sure my point in the previous comment is clear. I’ll try again. Assuming that all justification consists in evidence, the concept ‘defeater’ is needed only if you allow a *part* of one’s evidence to provide (propositional or doxastic) justification for P. If, strictly speaking, nothing besides one’s *total* evidence provides one with (propositional or doxastic) justification, then there is no need for ‘defeater’–unless we change the subject and talk about knowledge defeaters. Since it sounds like you only allow total evidence to provide justification, your view doesn’t have space for defeaters.

    Knowledge defeaters prevent JTB from constituting knowledge. “Justificational” defeaters prevent prima facie/pro tanto J from constituting ultima facie J.

  13. Chris,

    Right, I’m total evidence all the way. Now, it might be that only a portion of my total evidence is *relevant*. And it might be that only a portion of my evidence is in fact part of the causal explanation of why I believe. But that goes in the basing relation (where there’s some flexibility in “proper”).

    I don’t really like talk of knowledge defeaters either. I think it’s something going wrong in the basing relation that (at least typically) keeps JTB from being K.

    The closest I can come to making sense of the notion is that a proposition D would be a knowledge defeater for P (and, again, I think defeat in any event ought to be aimed at support relations) when my J for P essentially depends on D and D is false. But then it’s not the proposition that’s causing the problem, it’s my relying on it, and that’s a basing problem (I don’t have to be aware of it to be relying on it).

  14. It seems to me that they logic here goes something like this:

    1) S said p.
    2) S is generally reliable.
    3) from 2, one can assume S is reliable in p-type matters.
    5) because of 3, 1 gives support for p.

    It is later learned that S is unreliable in p-type matters i.e., that S’s general reliability does not extend to p-type matters. That is, the assumption in 3 is not correct.

    Since the assumption in 3 is not correct, 1 does not support p, because the support comes through the assumed reliability of S in p-type matters.

    I suppose one could say that what was defeated was the assumption of S’s p-type reliability based upon S’s general reliability.

    It also seems that learning that S in unreliable in specific p-type matters renders the general reliability irrelevant to the specific argument, since the role the general reliability plays in the argument is to support the assumption of reliability in specifically p-type matters.

  15. The reconstruction may well be something like that.

    But the short version of my point is that support relations hold between sets of propositions and hold necessarily if at all. One might say that it’s an essential property of [n/m of F’s I’ve observed have been G’s] that it support [n/m of all F’s are G’s] (or whatever example you please.

    Epistemic probability is relative to persons and defined over what sets of propositions constitute their evidence at a time. So if nec, p supports q and S has as total ev. p at t, then nec, S has support for q at t. Nothing is ever defeated.

    If one wants to stipulate that defeat talk just talk about change in what’s justified for one over time, then fine, but I don’t think this has been at all clear, and I think people have been mislead by the lack of clarity. We’d get something like this if we were clear (it would no doubt need some further chisholming).

    S’s J for p at t1 is defeated by d at t2 iff
    (i) At t1, S’s tot ev E supports p.
    (ii) at t2, S learns d.
    (iii) E&d do not support p.

    I’m not aware of anyone explicitly introducing the temporal indices into the definition, but I think it’s important. I claim that if I had time, I could find examples of where people were mislead by the lack of explicitness.

    You may think the point a small one, but the examples would prove otherwise! It’s essentially the same point Kagan makes. It seems to have been widely adopted in moral theory, so the folks over there must have thought it a substantive point. Here, I’m stumping for its adoption in epistemology.

  16. Hello epistemologists!
    I think it is of some help trying to analyze the property of being defeasible by means of defeasibiliy-types. We would answer the question ‘what is defeasibility?’with the following type descriptions:

    (DJ) There is defeasible justification or warrant. When S’s belief that p is warranted/justified by evidence E, this warrant/justification is defeasible when it could be decreased/lost by new evidence E’.
    (DR) There are also defeasible reasons. When S’s belief that p is warranted/justified by reason R, this reason is defeasible when there could be new evidence E’ such that, if E’ holds, then ~R.
    (DB) And still, defeasible beliefs. When S believes that p, this belief is defeasible when there could be an evidence E’ such that, if E’ holds, then ~R.

    To sum up: Φ is defeasible =df. It is possible that there is defeating evidence for Φ (where ‘Φ’ stands for warrant/justification, or evidence/reasons, or beliefs.)

    Let us pick up an example stressing (DJ). You have a red-table experience (E) which gives epistemic support for your belief that there is a red table there (p) – so that, p is warranted by E. Someone tell you that the table is in fact white, and that it was illuminated by red light (or you discover it by yourself). In this case, your defeasible justification is defeated by the new evidence E’ and, therefore, your belief that p is also defeated. However, your previous evidence E keeps on track. After all, you really had that experience; the evidence E is not defeated in this case.
    Now an example stressing (DR): you justificably believes that Tom is a logician (p), on the basis of the evidence that Tom is a philosopher and Every philosopher is a logician (E). Again, p is warranted by E. Professor Smart presents to you some philosophers who are not logicians (or you discover it by yourself). In this case, your defeasible evidence E is defeated by the new evidence E’. In fact, E’ entails ~E. Not only your evidence is defeated but also your belief which was supported by it. However, the epistemic support between E and P is not decreased or lost. In fact, E does still give warrant to the belief that p. But E no longer belongs to your belief system – so that you have no more reasons to believe p.
    Finally, an example stressing (DB): you believe that tomorrow you will be at the Annual Meeting of Epistemology (p), based on the evidence that you read about the meeting on the internet, your professor told you about it, etc. However, one of your friends from the Annual Meeting of Epistemology comittee calls you, saying that the event was postponed to next week. In this case, your belief p is defeated by the new evidence E’ – and E’ entails ~p. Once again, the epistemic support relation between E and P is not decreased or lost. Furthermore, E keeps untouched.

    I’m actually workin in the entailment relations between (DJ), (DR) and (DB).

  17. Luis, it sounds like we agree that there is no such thing as defeat, but we disagree about whether we should keep using the word.

Leave a Reply

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