‘I have reason to believe that I won’t be run over by a bus later this month’

Like most epistemologists, I find it intuitively unacceptable to say things like ‘I know that I won’t be run over by a bus later this month’. (How could I know that?)

Now, this intuition gives me a lot of trouble, because unfortunately for me, my various theoretical commitments make it hard for me to accept any of the familiar ways of accommodating this intuition. (Basically, these commitments seem to force me to say that this sentence will be true, in most ordinary contexts of utterance; so the only explanation that I can give for its apparent intuitive unacceptability is that uttering this sentence would be pragmatically misleading in some way.)

It’s just occurred to me that it also seems just as intuitively unacceptable to say ‘I have reason to believe that I won’t be run over by a bus later this month’. (Well, this might seem intuitively acceptable if I had some specific reason for believing this — such as that I knew that I would be spending all of the month trekking across a wilderness far away from any buses — but if my only reason for believing that I won’t be run over by a bus is that the probability of its happening is sufficiently low, then it seems intuitively unacceptable to utter this sentence.)

Do the other CD-ers share this intuition, I wonder? And if so, would those who hold that it is false, in most ordinary contexts of utterance, to say ‘I know that I won’t be run over by a bus later this month’ also think that the same goes for ‘I have reason to believe that I won’t be run over by a bus later this month’? How can we adapt the theories that imply that the former sentence is false in most ordinary contexts so that they also imply that the latter is false as well?


Comments

‘I have reason to believe that I won’t be run over by a bus later this month’ — 25 Comments

  1. It seems like a mistake to say that I won’t be run over by a bus this month or later this month. I would add that, insofar as belief is shaped by (putative) evidence, it seems wrong to say that I even believe that I will not be run over by a bus this month (or later this month). Of course, if we ignore flat-out beliefs, it may be less problematic to speak this way. So if we think about these matters along Bayesian lines the problem seems less acute (i.e., the problem of saying one believes one will not be run over by a bus or that one has reason for so believing). So if we’re talking about flat-out belief, and if we assume that beliefs are shaped by evidence in some way, given the slim evidence one has for the claim one is making about one’s belief, it seems false. I’m not sure if this speaks to the heart of the problem you raise or not. But I can’t think about it any more for now (I’ve got to go teach a class).

  2. Do you have similar intuitions if we modify Vogel’s “Heartbreaker” golf example? That is, if the question is not just whether you know that that you won’t be run over by a bus this month, but whether you know that all of your friends and relatives won’t be run over by a bus this month, do you have different intuitions? I confess that I do. Do I have reason to believe it? Well, no specific reason, I suppose. But I do think I’m reasonable in believing it. Anyway, my theoretical committments seem to entail that mere possibility of error won’t be enough to make knowledge assertions about the future false, but that sufficient possibility of error will be enough. And these entailments seem consistent with my different intuitions in those two cases.

  3. Hi Ralph, I share your judgments about the the intuitive oddness of the utterances. Yet I think they’re true. I’m pretty confident you do have sufficient reason to believe you won’t be run over… I think you also know it, though here I’m ready to be talked out of the thought.

    I don’t have any new ideas about how to explain the oddness. Just agreeing with your thoughts up to this point…

  4. My intuitions about pragmatics are roughly Gricean. If you say that you have reason to believe that you will not be hit by a bus later this month, your assertion is felicitous only if it is informative. That means you are implicating that you have reasons for your belief that go beyond those readily available to everyone or that go beyond the reasons your audience would assume you to have. Furthermore, felicity would require you to be in a position to say what reason you have for believing you will not be hit by a bus later this month, beyond what reason everyone has or your audience assumes you to have.

    Maybe ‘I have reason to believe I will not be hit by a bus later this month’ is true, but it looks false because its assertion would ordinarily be infelicitous. Its assertion would conversationally implicate various false things.

    Or maybe it’s an elusive claim: Although it is true most of the time, it is false in most of the situations where its assertion or denial are relevant to the conversation.

  5. I think this is exactly right: “[I]f my only reason for believing that I won’t be run over by a bus is that the probability of its happening is sufficiently low, then it seems intuitively unacceptable to utter this sentence.” In other words, a statistical or objective probability, without more, never is a reason to believe something. It never is rational to hold a belief simply on the basis of statistical evidence.

    If I had to gesture at a positive account, I guess I would say that reasonableness of belief has something to do with remoteness of possibilities (though I realize there’s nothing terribly original in that thought). Thus, I have a reason to believe (indeed, I presumably know) that I won’t be hit by a bus in the next five minutes, not because that event has such a low objective probability, but because it is so difficult to conceive of a way in which it could happen (I’m on the thirty-fifth floor of an office building and, alas, will remain here for sometime). In contrast, the possibility that I’m hit by a bus next week is not nearly so remote. Unfortunately, this only minimally advances the ball.

  6. First, unlike Ralph and Jim Pryor, I do not think it sounds bad to say that I have reason to believe I won’t be run over by a bus. I think I do: very few people get run over by busses! (Likewise, I have reason to believe I won’t win the lottery tomorrow, even though I know I’ve bought a ticket.) By contrast, it does sound bad to say that I know I won’t be run over by a bus.

    Second, I cannot believe that Stephen Fogdall’s suggestion is correct:

    “It never is rational to hold a belief simply on the basis of statistical evidence.”

    I know that no team with the Packers’ current record (1-6) has ever made it to the Superbowl. This is statistical evidence (at least I think it is!), and my possessing it seems to me quite sufficient to make it rational for me to believe that the Packers won’t go to the Superbowl this year.

  7. Fwiw, my intuitions are in complete harmony with Geoff’s. My intuitions are that I’ve got reasons; but that I don’t know; and that very often, though not always, statistical evidence does a fine job of granting us knowledge. I guess I’ll add that some of the machinery in Hawthorne’s lotteries book does seem to me to go a long way towards distinguishing the cases in which epistemologists will have positive or negative epistemic intuitions in statistics-type cases.

  8. Hi Ralph,

    Is it something special about the future, do you think, that guides your intuition in this case? I’m curious if you have the same intuition about “I know that no one in my extended family has been run over by a bus in the last hour”. (Suppose that your extended family contains 720 people — so the odds of one of them getting run over by a bus in the last hour are roughly equal to odds of your getting run over by a bus within a month, supposing the last hour to be representative.)

  9. I certainly don’t think it sounds false to say that you have good reason to believe you won’t be run over by a bus this month. And I’ve yet to be able to get myself to hear it as infelicitous.

  10. My non-epistemologist intuitions are that there is nothing wrong with saying I know I won’t get run over by a bus this month. It’s true, I think, and I imply nothing misleading by saying it. (This last bit is theory and not intuition.)

    I do think saying I have evidence is odd. But my guess as to its oddity is that this is explained by it being weaker than other true things we can say. Why only say there is evidence or that I have evidence when I know? (Most of this is theory, but the theory makes me feel like I understand my intuition.)

    For what the intuitions of non-epistemologists are worth . . .

  11. I don’t think it just sounds odd to say “I know I won’t get hit by a bus later this month,” It’s straight forwardly false. This is because, in my view, there is an easily forseeable situation in which the matter will be unambiguously settled, and you are not yet in that situation. Any bets on whether you will get hit by a bus would naturally be settled at the end of the month and not before. Suppose, rather sinisterly, you stumbled across a couple of people who had a bet on whether you would get hit by a bus within the month. You would be in no position to settle their bet for them. Whereas if you knew you weren’t going to get hit by a bus, then you could settle it for them.
    “I have reason to believe I won’t get hit by a bus this month” on the other hand is true, and the pragmatic explanation for why it sounds odd is clearly the right one. You seem to be implying that there is a special reason for you to believe that, not just information available to everyone.

  12. It has always seemed to me that, at least when it comes to beliefs about contingent physical events, reasons depend on physically plausible (or at least possible) causal or explanatory relationships. Geoff, using your Packers example, the statistical evidence you mention, standing alone, doesn’t provide any envisionable causal or explanatory story as to why the Packers won’t win the Super Bowl. To the extent you feel that you do have reasons to believe this, it seems to me that you implicitly are drawing on other evidence more directly tied to specific facts about the Packers and the other teams in the league. The statistical evidence in essence is standing as a surrogate.

  13. Someone once wrote that in situations where it’s unclear whether a given claim is false or merely inappropriate, it’s a good idea to check the negation of the claim in question. Sometimes following that methodological advice has quite definitive results. This isn’t one of those cases. Still, it’s worth asking. How do those who find “I have reason to believe that I won’t be run over by a bus later this month” feel about “I don’t have reason to believe that I won’t be run over by a bus later this month”? Would be helpful to know. If you find them both somehow wrong, that’s good reason to think there’s something pragmatic (in the needed use) going on here. If on the other hand only “I have reason…” seems wrong, and “I don’t have reason…” seems perfectly fine, a pragmatic treatment is not nearly as likely to be what you need.

    (If you “check the negation” of the knowledge claim, it’s important to try it without emphasizing the “know”. Instead of “I don’t know that I won’t be hit,” try saying it in a “flat-footed” manner.)

    –One of the remaining practioners of that dying ancient religion

  14. Forgot a couple of words. Should have been:

    How do those who find “I have reason to believe that I won’t be run over by a bus later this month” somehow wrong feel about “I don’t have reason to believe that I won’t be run over by a bus later this month”?

  15. I’d toss my markers in with Jonathan and Geoff, too. There are tricky issues involved in moving from a typical case (e.g., are we to sample from the population of pedestrians; sample from pedestrians in a US city; sample from pedestrians in Lisbon, Portugal (God help you!) and so on) in order to arrive at an estimate, a reasonable estimate, an estimate that is reasonable to accept as representative of the particular case one finds himself in (e.g., death by bus, given my daily walks, over the next 30 days, to and from the Avenida Metro station in central Lisbon). But, it seems to me that we do this. And quite often do it well.

    There are plenty of wrinkles here, many of them very interesting. One of them is the roll that background information plays, which is the point that Stephen is hammering on. (BTW, Stephen, would you kindly send me an e-mail? I lost your e-mail address in ill-planned experiment this spring…). But I think his conclusion is too strong. For example, Henry Kyburg certainly (!) endorses the view that statistical statements license knowledge claims, yet his system for generating a minimal cover, one that fixes lower and upper probability on the event in question given all the information one knows about the particular case, is determined both by one’s total knowledge but also by general principles for sorting out conflicting statistical statements. Since our knowledge is incomplete, statistical formulas over generate (a.k.a. the reference class problem).

    So, of course statistical statements don’t come out of thin air. But neither do probabilities. The “remoteness of possibilities” condition that Stephen mentions is part of the preconditions for doing statistics, for applying probability, for applying mathematics in general, I dare say. In this case, delimiting the set of outcomes to just that set thought to represent what is possible is a necessary condition to conceive a sample space for your measure, whatever framework you settle on using.

    This said, I’m not sure how to get Stephen’s argument to work here. I’m not seeing how to go from the claim that some background knowledge is necessary to generate a statistical statement to the conclusion that (to paraphrase) statistical statements are neither a sufficient basis for knowledge nor rational belief.

  16. My intuitions: It is true that I have reason to believe I won’t be hit by a bus, and false that I know I won’t be hit by a bus (pretty much). But I find it infelicitous to say “I have reason to believe I won’t be hit by a bus,” at least in most conversational contexts I can think of. (Not in this one, of course! I just said it.)

    And I think Chase diagnosed it correctly in 4: In most contexts there are Gricean reasons not to say it. Since everyone knows that it’s comparatively rare to be hit by a bus, if I say that I have reason to believe I implicate that I have an additional reason to believe beyond the one everyone already knows I have. (Except in this context, where we aren’t presupposing that the probabilities give me reason to believe.)—I’d say something similar about the infelicity in many contexts of saying “I won’t be hit by a bus.”

    As for the negated case, I think it sounds odd to say “I don’t have reason to believe I won’t be hit by a bus.” If someone said that to me, I’d respond “Are many people round here hit by buses?” But I think “I don’t have any particular reason to believe I won’t be hit by a bus” sounds fine–maybe with stress on “particular” or “I”—in that it suggests that I have no reason beyond the statistics to think so.

    The Gricean explanation predicts that, when the statistical information isn’t widely known, it will be felicitous to say “I have reason to believe that I won’t be hit by a bus.” And I think this checks out. Suppose I’m about to ship out to a war zone, and someone says, “God, of all the things that can happen to you, you might get hit by a bus as well.” I say, “I have reason to believe that won’t happen—only two of 100.000 GIs in the zone were hit by buses last year.” Isn’t this OK? I don’t have any particular causal story, I’m just relying on probability.

    But I can’t toss my marker in with Geoff completely, because I do think it sounds odd to say “I have reason to believe…” in most contexts when the reason is based on widely available statistical evidence.

  17. What widely available statistical evidence? I tried to look up the statistics on getting hit on a bicycle, and found out that sober drivers are just as likely to crash as drunk drivers, I have never got anyone to believe this, or to think that I have a reason to believe it, in fact people get quite angry (and don’t tell me that there are more sober drivers, I checked the base rate). If I had just looked up the statistics on bus death, then I would have a reason to believe that I wouldn’t get hit in the next month. If I just said it out of the blue, I would be making up the statistics to support my unjustified belief, not the other way around.

  18. Oops, ‘statistical’ was misspeaking. What I meant was, I and I presume everyone else have good evidence that being hit by a bus is rare, and I think that gives us reason to believe that we won’t be hit by buses in the next month (and that this reason doesn’t seem to qualify for what Stephen said about causal/explanatory statements). that evidence: I have never been hit by a bus, none of my acquaintances has, it’s very rare that I read about someone being hit by a bus and if it happened often I’d expect to—that sort of thing.

    There are interesting questions about how the Gricean analysis is affected if we don’t in fact have good reason to think that being hit by a bus is rare—it maystill go through—but I’ll have to get back to that.

  19. Matt, I think what you have is a reason (a strong reason) to accept the proposition that you probably won’t get hit by a bus, not a reason to accept the proposition that you won’t get hit by a bus. That proposition isn’t detachable from its probabilistic underpinnings here. You can detach it when the possibility becomes suitably remote. But remoteness requires more than improbability. Remoteness requires some reference to the physical character of the situation.

  20. I think this may just be an intuition clash here—or a difference between belief and acceptance—but while I agree that any reason to believe “I won’t be hit by a bus” comes from a reason to believe “I very probably won’t be hit by a bus,” I don’t have a problem with saying that I have a reason to believe the former proposition. That’s supported by my intuition about the war zone case.

    (Note that I also think that, in a suitably de-Griced setting, it’s OK to assert “I won’t get hit by a bus,” and many people would read me out of the movement for that!)

  21. Stephen, I am interested to find out what is meant by remoteness? Probability is subjective when it is relative to a set of beliefs or assumptions. Objective probability I never really understood, but it is relative presumably to a set of prior states of affairs. Remoteness, I guess, is the distance of the possible world from the actual world? Has this got anything to with probability either objective or subjective? If they are completely different, this would explain why we can know things without eliminating all possibilities, whereas we don’t know things which are highly probable.

  22. Matt wrote: Since everyone knows that it’s comparatively rare to be hit by a bus, if I say that I have reason to believe I implicate that I have an additional reason to believe beyond the one everyone already knows I have.

    I think this is only 1/2 right. Whether one would implicate that depends on specific features of the conversation, too. Compare:

    You are offering me bus-crash insurance that pays ten million dollars if I get hit by a bus, and it costs a penny. I say ‘I won’t get hit’/’I know I won’t get hit’/’I have reason to believe I won’t get hit’/etc.

    These are appropriate only if I have evidence that warrants not taking the deal, for the Gricean reasons mentioned.

    I think ‘I don’t have reason to think I won’t get hit’ passes Keith’s test – that sounds wrong, and the Gricean can just say: wrong, because false.

    I find knowledge denials sound right only when the stress is on ‘knows’. ‘I don’t know that I have hands’ sounds fairly natural, but ‘I don’t know that have hands’ or ‘I don’t know that I have hands’, to my ear, would only be appropriate if fake-hand replacements had been discovered in a large portion of the population, and I hadn’t had my “hands” checked yet.

    It also seems like one could say ‘I don’t have reason to think I won’t get hit’ and get away with it, at least in the right context. But Keith, is your point that the pragmatic account is in trouble only if the unstressed versions (without stress on ‘know’ or ‘reason’) are acceptable? I agree with that … but I think the unstressed versions aren’t acceptable …

  23. Jonny, re your post 21, this intuitive idea of remoteness presumably could be understood by comparing possible worlds, but my suspicion is that it couldn’t be captured by a distance metric. A more qualitative comparison would seem to be needed. Perhaps a relation of “feasible accessibility,” or some such thing. The literature on relevant alternatives might also shed some light here. As to your question whether this has anything to do with subjective or objective probability, I assume that if there is a workable notion of remoteness (or feasible accessibility) among possibilities, this notion also could play a role in a propensity account of single-case probabilities, which I view as a type of objective probability. But the probabilities would be analyzed in terms of the feasible accessibility relation rather than the other way round.

  24. Thanks, I’m looking into relevance theory and logics and I’m not sure whether this is related to relevant alternatives. I geuss remoteness could be measured in the amount of effort in cognising the possibility. Psychological studies have indicated that people may match their expectations to the ease of imagining, and this is set by the amount of similar memories. How easy is it to imagine that you will get hit by a bus? It actually seems fairly easy, given that you cross roads every day. This is why having a good reason to think that you won’t get hit by a bus requires something stronger than the mere background knowledge that it doesn’t happen that often.

  25. Pingback: Certain Doubts » Stats for November

Leave a Reply

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