Klein on Useful Falsehoods

Peter Klein was just here and gave a stimulating talk on useful falsehoods. The idea is that sometimes you know things even though the evidential path relies essentially on falsehoods. Here’s one of his examples:

You believe that your secretary told you on Friday that you have an appointment on Monday. As a result, you believe that you have an appointment on Monday, and since your secretary is reliable, trustworthy, etc., and what she told you is true, you know that you have an appointment on Monday. But she didn’t tell you this on Friday; she told it to you on Thursday.

The standard defeasibility account Peter defends does not accommodate this example, so Peter’s goal is to characterize useful falsehoods in order to adjust his undefeated justified true belief account of knowledge.

The account he develops takes it’s cue from a remark by Hilpinen that you can acquire knowledge through falsehoods if your false belief is close enough to the truth. Here’s the full account:

The belief that uf is a useful falsehood to S (for acquiring knowledge) by producing a
doxastically justified belief that h iff:
1. uf is false
2. The belief that uf is doxastically justified for S
3. The belief that uf is essential in the causal production of the belief that h
4. uf propositionally justifies h
5. uf entails a true proposition, t
6. t propositionally justifies h
7. Whatever doxastically justifies the belief that uf for S also propositionally
justifies t for S.

The clause that makes this an elaboration of Hilpinen’s remark is clause 5, that the false belief has to entail a true one, where the notion of entailment is something like the relevance notion of it.

There is an alternative, though, and neither Peter nor I can think of a reason that it won’t work. What matters in the secretary case is that you could have come to the conclusion that you have an appointment on Monday by reasoning through the claim that your secretary told you sometime during the past week that you have an appointment on Monday. This claim is true, and is related to the false belief that you actually reasoned through in two different ways. First, the false belief entails the true proposition in question. Second, the false belief is a reason to believe the true proposition (and the reasons for believing the false belief are equally good reasons for believing this true proposition).

So there are two different clause 5’s that could be proposed. There is the alethic version espoused by Hilpinen and clarified in terms of the notion of relevant entailment by Peter. There is also the epistemic version, according to which what matters for that particular clause is just the idea that your false belief propositionally justifies the true proposition in question.

Suppose we call the first view the alethic presupposition view (since x presupposes y in this sense only if x entails y). And suppose we call the second view the epistemic presupposition view. The question then is this: is there any reason to prefer either view to the other in accounting for cases like the secretary case?


Klein on Useful Falsehoods — 33 Comments

  1. The 2nd view handles more cases than the first — cases that need to be handled. But both views have problems. I’ve given a few talks at places in the last year on this stuff and I’m going to give a few more before publishing (or trying to publish) a paper about them. I’ll share the paper when it’s ready.


  2. I was thinking about how Kleinâ??s account of useful falsehoods fares with Gettier cases and I thought up the following counterexample. This is a case in which all the conditions of a useful falsehood are met, but it doesnâ??t help S gain knowledge.

    A variant on a Gettier case: I have lots of evidence that Jones owns a Ford. I believe that Jones owns a Ford. I extend rational belief by implication to either Jones owns a Ford or Smith is in Barcelona. I then further extend rational belief by implication to some disjunction is true. The causal basis for my belief that some disjunction is true is my belief that Jones owns a Ford. But I donâ??t know on that basis that some disjunction is true. Hence the belief that Jones owns a Ford is not useful for me in gaining knowledge that some disjunction is true even though it satisfies all the conditions 1-7.

    In my case â??ufâ?? is Jones owns a Ford. Itâ??s false and doxastically justified for S. â??hâ?? is some disjunction is true. Condition (3) is satisfied because uf is essential in the causal production of the belief that h. Condition (4) is satisfied; uf propositionally justifies h. Condition (5) is satisfies because uf entails either Jones owns a Ford or Smith is in Barcelona, which is â??tâ??. (6) is satisfied because t entails h. And (7) is satisfied.

  3. Ted, I think your first paragraph shows a misunderstanding of Klein’s view. I think the view he holds is that some cases involving useful falsehoods count as knowledge and some don’t. Those that don’t will be explained by his standard defeasibility theory, and your case is one of those. So I don’t think your example is a counterexample to his theory.

  4. Jon, I don’t see how an appeal to the defeasibility theory will help in this case. Klein’s account of inferential knowledge in the paper is as follows: S inferentially knows that h iff (i) h is true; (ii) S believes that h; (iii) S’s belief that h is inferred from and doxastically justified by another belief, say the belief that e, which is doxastically justified; and (iv) There is no genuine defeater of the propositional justification of h by e.

    As he pointed out in the talk condition (iv) needs to be modified to allow for useful falsehoods because in the case where e is a useful falsehood ~e is a genuine defeater to the propositional justification of h by e. If this is right, then I would expect condition (iv) to be reformulated to restrict genuine defeaters to those that aren’t useful falsehoods. But in that case I think my counterexample holds.

  5. Ted, I think what you suggest for modifying clause (iv) Peter explicitly denies. Consider:

    It is important to note that this definition does not yet fully explicate the role of false beliefs in the production of knowledge. Here we are concerned only with the so-called third condition in the analysis of knowledge � the doxastic justification condition. The definition is intended to explicate the conditions under which a false belief can produce a belief which is �held for the right reasons� � reasons which, ceteris paribus, are sufficient to bring about knowledge. A central notion is that even though uf is false, it is �close enough to the truth� so that if believing it causes S to believe that h, then, ceteris paribus, S knows that h. . . . Indeed, as we will see, all that is required to omplete the task of explicating the role that false beliefs play in the production of knowledge is to flesh out the ceteris paribus clause by requiring that there is no genuine defeater of the evidence path from x to h that goes through t.

    Moreover, if I look back at your case, I think you want the following values for uf, t and h.
    uf=jones owns a ford
    t=the jones/smith disjunction
    h=some disjunction is true.
    But this chain of reasoning is clearly defeated, since uf is false. What you need is a value for t that provides an alternative path to the one actually taken. Now suppose you suggest the evidence for uf itself, so that you go straight from that evidence to the jones/smith disjunction. I doubt that the propositional justification for uf is, by itself, a propositional justification for this disjunction.

  6. Jon, I’ll have to think more about the modification of (iv). If my revised Gettier case meets all the conditions of the account of useful falsehoods, what other condition fails–besides the fact that uf is false–that does not make it a case of K?
    About my counterexample, those are the values I want for uf, t, and h (the transition from Word to the web garbled it all up). To get propositional justification for t by not going through uf, there could be another line of evidence for t by way of evidence for Smith is in Barcelona. Perhaps it’s a deductive consequence of my beliefs that Smith is in Barcelona. It’s a very complex deduction that I haven’t performed but it provides propositional justification, not doxastic justification, for t.

  7. Ted, you say,

    To get propositional justification for t by not going through uf, there could be another line of evidence for t by way of evidence for Smith is in Barcelona. Perhaps it�s a deductive consequence of my beliefs that Smith is in Barcelona.

    That’s not Klein’s account. Look at clause 7; whatever doxastically justifies uf has to propositionally justify t.

  8. OK, I agree. I’ll take the more direct option going through uf. In the original case I have lots of evidence that Jones owns a Ford. The evidence propositionally justifies that Jones owns a Ford. If propositional justification is closed under entailment then it propositionally justifies t. Since I’m doxastically justified in believing that Jones owns a Ford then t is propositionally justified for me. So it seems that condition 7 is met.

  9. Sorry, Ted, I think it still won’t work, and I think you know why: propositional justification is not closed under entailment. That’s why Klein requires that the notion of entailment be interpreted as relevant entailment, and of course, one the features of relevance logic is that you don’t get wedge-introduction.

  10. Peter might be doing something different than what I’m trying to do in working on useful falsehoods. If you’re right Jon, Peter isn’t trying to distinguish “knowledge from falsehood” cases from “no knowledge because falsehood” cases. This would be unfortunate because that’s where the serious action is at. Are you sure you’re right that Peter’s “useful falsehoods” might are only useful (for knowledge) up to the point of Gettier issues? (but silent about the distinction I note above between “knowledge from” and “no knowledge because…”?)? Textually, I’m guessing you are right because Peter’s account won’t distinguish useful-for-knowledge cases from Gettiered cases and Peter would surely know this.

  11. Fritz–yes, Peter’s project is not to say when and only when knowledge can rest on false beliefs. His definition of a useful falsehood still leaves open that you can have a useful falsehood coupled with a true belief and still lack knowledge. Plus, if I remember correctly, he also thinks there are cases of knowledge resting on false beliefs, where the falsehoods in question are not useful falsehoods (they are benign falsehoods instead).

  12. Interesting. Of course by “benign” falsehoods I assume we all mean “falsehoods not essential to reaching the conclusion” (irrelevant). So like this example:

    Kvanvig is a living epistemologist.
    Klein is a living epist….
    Warfield is a living epist….
    Chisholm is a living epist….
    Foley is a living epist….
    So, there are at least two living epistemologists.

    I know my conclusion despite the falsehood in my reasoning — it’s an irrelevant falsehood. Much more interesting are cases in which the falsehood involved appears to be essential to the inferential argument in question — perhaps it appears to be a one material premise modus ponens argument. (though “essential” never seems to me to be the right word here….)

    Here I think the interesting project is to try to distinguish cases of “knowledge from falsehood” from cases of “no knowledge because of falsehood” and that enters Gettier territory. I confess to not full understanding what this other project would be — unless it’s the same project but stopping short of dealing with the Gettier issues? strange.

  13. Fritz, I was worried along the same lines when I first saw Peter’s paper, Peter addresses this issue in the paper as well. The idea is to take on those who claim that essential falsehoods automatically bar one from knowing, and the concept of a useful falsehood is supposed to undermine that view. Given that goal, one could try to define a useful falsehood so that every case involving such a falsehood is a case of knowledge, or one could take Peter’s route and define the notion so that it can be used to argue against the standing orthodoxy. Ultimately, though, we want what you are trying to do: an account that tells us in exactly which cases knowledge and falsehoods coexist.

  14. Peter’s project has this advantage over mine: mine is much harder than his! Shouldn’t Peter have to do the hard project? Maybe he and I can trade….

  15. Jon,

    Two of the claims you´ve made above about how Peter Klein thinks of propositional justification strike me as uncharacteristic of him. And the context doesn´t make it clear whether you´re _reporting_ what you may have heard from him or whether you´re offering your own interpretation of those points in his work on useful falsehoods. Maybe you can clarify. These are the troubling points:

    1. You suggest that Peter´s interest in relevant entailment may have something to do with prohibiting wedge introduction. I´d be surprised if that´s what he thinks. I suggest that what matters for him in relevant entailment is simply the twin ideas that contradictions don´t justify anything and tautologies are not justified by everything. (I speculate.)

    2. Peter has famously argued that propositional justification is closed under entailment. And he has also famously argued that prop. justification is not transitive. Is it at all conceivable that you were thinking of the latter when you suggested that he may endorse the denial of closure? Or were you expressing your own personal view regarding closure?

  16. Claudio, good to hear from you! On the first question, Peter mentioned in discussion the need to interpret his notion of entailment as relevant entailment, and it is distinctive of that approach to block wedge introduction. But I don’t recall him mentioning that particular rule, so that part is interpretation. I haven’t seen what Peter says about justification being closed under entailment, but I’d be shocked if he holds that view. In the context above, remember that being closed under entailment means: if you’ve got PJ for p, and p (classically) entails q, then you’ve got PJ for q. That claim is so obviously false, and Peter is so obviously a very good epistemologist that…

  17. Hey, Jon, you surely know how to make this a lively place! Thanks!

    Now, I´m sure you have noticed that I didn´t say that Peter endorses the principle that, if you have PJ for p, and p entails q, then you have PJ for q _without qualification_. One of the restrictions he has insisted upon is that, for that version of the closure principle to be true, p and q must be contingent propositions. Otherwise, the principle is obviously false (since q could be a necessary truth). I thought that was clearly implied in my note above. Clearly, however, it doesn´t seem to have been.

  18. Claudio, you’re right that the principle is better if restricted to contingent claims, but Peter was clear that he wanted further restrictions. The entailment, even for contingent claims, has to be a relevant entailment.

  19. Yes, Jon, he does want other restrictions. Here´s a passage from his “Skepticism” (Oxford Handbook of Epistemology, 2002):

    “But CP [the closure principle for propositional justification] can easily be repaired. We can stipulate that the domain of the propositions in the generalization of CP includes only contingent propositions that are within S´s capacity to grasp and that the entailment is ‘obvious’ to S.”

    I´m not buying the additional restrictions, though. When it comes to _propositional justification_ — as opposed to doxastic justification — I can´t say that I understand why propositions which are beyond S´s capacity to grasp and non-obvious entailment is a problem. I see no problem with the view that you´re entitled to believe a whole lot more than you´re currently dim lights will allow you to believe with doxastic justification. Propositional justification is just a necessary condition for doxastic justification.

    In any case, that´s Peter. The paper in the Oxford Handbook is a good instance of his defense of closure for propositional justification. In “Closure Matters: Academic Skepticism and Easy Knowledge”, a talk he gave at the SOFIA conference (my university, PUCRS, Brazil, May 2004), forthcoming in Philosophical Issues, he reasserts his commitment to the qualified version of the closure principle for PJ. There´s also his “Skepticism and Closure: Why the Evil Genius Argument Fails” (Philosophical Topics, 1995), his book and maybe other papers that I forget now.

    As regards “relevant entailment”, again, I doubt that he´d subscribe to any relevant logic that prohibits wedge introduction. Maybe the talk of “relevant entailment” is a misnomer for what he´s after. But I´m just speculating, of course. One thing seems clear enough: If all you want to say is that it is false that necessary truths are propositionally justified by everything (for a given agent), you don´t need relevant logic as a part of your case for a properly qualified closure principle.

  20. Claudio, excellent! I’ve posted here about the problem of propositions you can’t conceive, which is a worse problem of the two Klein mentions. But you’re right that there’s no reason to limit the closure principle to obvious entailments. On the other issue, I’m inclined to think that there are contingent propositions entailed by our present evidence that are beyond our capacity to grasp and not propositionally justified for us because of that fact. But I don’t have an argument, other than through very general features concerning the perspectival character of justification, which ought to be true not only of doxastic justification but propositional justification as well.

    Peter was clear while here, though, in his use of the language of relevant entailment. I wonder if he’s changed his views a bit…

  21. Thanks, John. Yes, like you, I don´t see a good argument in the offing to the effect that your conceptual/inferential abilities put a cap on what is propositionally justified for you (an infinite number of propositions, if you go with closure). Unlike you, on the other hand, I have no tendency to think that we should be looking for that argument. All is well on that front, I think.

  22. Claudio, how about this for the beginning of an argument? A closure principle ought not be limited to contingent propositions, since it’s obvious that a primary home for application of closure is in math and logic. An acceptable and fully general closure principle is going to have to be restricted in terms of the capacities of the person in question to avoid absurd implications, and if we accept such restrictions in the math/logic domain, we should accept them for contingent claims as well.

  23. Huge topic, Jon! But I wouldn´t do it your way. (Maybe I misunderstand what you suggest. In any case, this is the place to philosophize dangerously, isn´t it?) Some of the ideas I´m putting forward in a forthcoming paper _may_ give us the makings of a solution. I hear that the Certain Doubts crowd is _very_ charitable; so, here´s the quick-and-dirty version:

    Forget the classical story about logical implication (or validity). It´s obviously oversimplified. It´s laughable to the unpolluted minds of our undergraduate students! What else do we need?! But that doesn´t mean we´re giving up on classical logic. The fact that it´s obviously absurd to claim that â??Mars is a planetâ?? implies â??2+2=4â?? just because it is impossible for the implied proposition to be false and there´s a story about logical implication according to which blah-blah-blah doesn´t mean you´re now forced to give up on classical rules of inference which seem unimpeacheable to you (unless you´re an anti-realist, for instance, or a relevant logician). It means only that we refuse to avail ourselves of the classical story about validity. It MUST be wrong to think that, if the classical theory of validity goes to hell, classical rules must go with it. (I think this is implied by something David Sanford once wrote about classical validity.)

    Now, do keep closure fully general, as you suggest, and take â??p entails qâ?? to mean roughly the same as â??there is a deductively unobjectionable inferential pathway leading from p to qâ??. Plug your favorite logic into that reading of â??p entails qâ??, and you´ll have no need for a reference to conceptual/inferential capacities as part of an account of propositional justification. This way, the notion of propositional justification is compatible with my previous claim: It seems entirely natural to me to think that you´re entitled to believing a whole lot more (infinitely more) than your dim lights presently allow you to believe.

    (For now, I ignore the part about prohibiting the propositional justification of everything by a necessary falsehood. Maybe later.)

    Can you use this?

  24. Well, it is nice to hear we have a reputation here for being charitable! I do like the pre-theoretical account of validity that you employ, and you’re certainly right that a rejection of classical logic doesn’t require a rejection of all classical rwules. I’m also untroubled by infinities in epistemology: either the claim that we have infinitely many (dispositional) beliefs or that an infinity of claims are justified for a person. The part I remain uncertain of, though, is at the end of your suggestion. No matter what logic one employs, there will be unobjectionable inferential pathways from math/logic premises to math/logic conclusions where the conclusion is inconceivable to the person in question.

    But maybe you want the inferential pathways to be conceived of in terms of the relata and not just the relation? In such a case, the inferential pathway would be unobjectionable only in relation to the values of p and q, and perhaps we could say that the unobjectionability couldn’t obtain without p and q being within the conceptual capacities of the person in question. That would handle my concerns here, but I think it won’t do justice to the idea of a logic, which seems to me to be centrally about characterizing in the metalanguage rules of inference with unrestricted metavariables for any wff whatsoever in the object language. Of course, that is no real objection to specifying a notion of entailment that will avoid certain problems with the standard, logical notion, but it does make it hard to rely on any of the usual logical systems or the motivations we have in constructing them.

    I would think, though, that the issue of inconceivable propositions doesn’t worry you as much as I, and that’s in part because I haven’t given any compelling reason to take it as seriously as I do. Another way to put the issue is that I want more perspectivality in propositional justification than you may be inclined towards, and I don’t know how to argue for that except to appeal to symmetry considerations regarding doxastic justification that are not even close to being compelling or even very persuasive.

  25. No one will be reading this any more, but I’m not convinced by this case of useful falsehood. Here’s an alternative account:
    You dispositionally believe (and know) that the secretary told you you have an appointment on Monday (full stop). This is true and is the basis for your knowledge that you have an appointment on Monday. And it is not based on the proposition that she told you “on Friday” that you have an appointment on Monday, either. Rather, it would seem to be justified by the experiential memory of being told by the secretary, etc.

  26. Mike, I hereby refute you! Well, the “no one” claim, at least…

    Do you think a person can believe that their secretary told them X on Friday, but not believe that their secretar told them X? I don’t see why that’s impossible, and if it is possible, then if that possibility is realized, the basing will go through a false belief.

  27. Jon,

    I don’t know that that’s impossible, but I don’t see how it is; someone would have to tell me a story about how one could do that.

    Again, we’re looking at dispositional belief here. What are the conditions for dispositionally believing that P? I don’t know exactly, but I think it is a matter of having an occurrent belief at some time, which is appropriately related to P, one that we could say P is (at least) “implicit in”. My thought as to why you couldn’t believe that “S told me P on Friday” without believing “S told me P” is that the latter is implicit in the former. Compare a related case: It seems that one cannot believe that (P & Q), without believing that P.

    Also, in the example, you’re going to have a memory experience representing the secretary telling you P, and you’re going to have some sort of implicit trust in the memory. Also, your dispositions are going to be as if you believe the secretary told you that P. For instance, if asked, “Hey, did the secretary tell you you had an appointment on Monday?” you’ll say “Yes.” If some practical decision turns on whether the secretary told you that P or not, you’ll act as if she told you that P. Again, someone would have to tell me a story to show how you might fail to believe that she told you that P. Then we’d have to look at that case and see whether we still think you know that P.

    A more promising avenue, I think, would be to say that your belief that she told you that P is based essentially on the belief that she told you that P on Friday. I’m going to say that in order for that to be the case,

    (a) the justification you have for “she told me P on Friday” has to fail to independently support “she told me P”; in the case of justification through memory, I think the evidence would support the latter just as directly as the former. Or,
    (b) the fact that the event was on *Friday* has to make a difference to what you believe.

    So here’s a story: you think that you have some sort of weird memory quirk that causes your memories formed on Fridays to be much more reliable than memories formed on other days of the week. Consequently, you would not believe that the secretary told you that P, unless you believed that she told you P on Friday. Here, I think your belief that you have an appointment on Monday is based on your belief that the secretary said this on Friday, and is not independently supported by the belief merely that the secretary told you this (at some time or other). But in this case, it seems pretty clear to me that you do not know you have the appointment on Monday.

  28. Mike, what I was thinking of was the difference between a disposition to believe and act in accord with the truth of a claim, and dispositionally believing that claim and thereby being disposed to act in accord with it. This distinction comes from Audi, and there seems to be one, though I suspect it is an undetectable distinction.

  29. That’s a good distinction. But at least the disposition to accept and act in accord with a claim is usually an indicator of the having of a dispositional belief, even if it doesn’t guarantee it. And again, I think it’s important that one has, by stipulation, an occurrent belief whose content is very close to that of the putative dispositional belief–that makes the case for one’s having a dispositional belief here pretty strong. Even though I don’t have an account of the conditions for having a dispositional belief, this seems to me like about as good of an example of such as I can think of.

  30. Mike, yes, I find the conjunction example particularly plausible, and it’s a really good reason to think that the person believes both that they were told and that they were told on Friday.

  31. Pingback: Certain Doubts » Testimony, Lies, Fiction and Benign Falsehoods

Leave a Reply

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