Mike Leigh on IRI

Jerry Dworkin pointed out to me that Mike Leigh’s film “All or Nothing” contains a scene that seems to support interest-relative-invariantism. The script is here, though be warned that link contains pop-ups.

Husband: Give us a clue, then.
Wife: ‘Biblical son of Isaac, five letters.’ Starting with a ‘J.’
H: Jonah.
W: Oh, yeah.
H: No, it ain’t. It’s what’s-his-name. Jacob.
W: Are you sure?
H: Yeah.
W: It’s a thousand pound prize.
H: Is it? No, I ain’t sure, then.

Well, maybe it is only interest-relative-invariantism about ‘sure’, rather than ‘knows’, but it seemed like a good way to celebrate the publishing of Jason Stanley’s book on IRI. More on IRI after the fold.

My version of IRI was based on the following two idea. S has a justified belief in p if her degree of belief in p is justified, and that degree is high enough to amount to a belief. Since what degree of belief S must have in p to count as believing that p is sensitive to S’s context (or interests or whatever) this is something like a form of IRI about justified belief.

Now it would be nice if that ‘if’ in the theory could be strengthened to an ‘iff’, but I don’t think that’s so, because of the following two cases.

Confident Carla and Sceptical Suzy each have the same evidence e, and each are in the same practical situation with respect to p. In that situation, degree of belief 0.8 suffices for belief in p. That is, they don’t face any choices that amount to bets on p at odds greater than 4 to 1 against.

  • The evidence e justifies a degree of belief in p of 0.9.
  • Confident Carla is too confident about the strength of the evidence; her degree of belief in p is 0.95.
  • Sceptical Suzy is too sceptical about the strength of the evidence; her degree of belief in p is 0.85.

Question: Which of the two of them justifiably believe that p? I’m inclined to say that they both do, though I’m more confident about Suzy than about Carla. But the little theory above, if strengthened to a biconditional, would say that neither justifiably believe that p. That seems to be a problem.

I think the thing to say is that S justifiably believes that p iff S believes that p, and S is justified in having a high enough degree of belief in p to count, in her context, as believing that p. That makes both Suzy and Carla justified believers, though it is a more complicated theory than the one I hoped would be true.

6 Replies to “Mike Leigh on IRI”

  1. It’s a lovely bit of dialogue from Mike Leigh there, but clearly it only supports an interest-relative position about being “sure”. I take it that being “sure” about a proposition p is a state that can be defined in purely psychological terms. Roughly, it is a state in which one simply takes p as a premiss in one’s further reasoning, including practical reasoning; in effect, one does not really go to the trouble of attaching any real credence to any propositions incompatible with p; one reasons simply on the basis of p — instead of reasoning both on the basis of p and also on the basis of various incompatible propositions, and weighting the conclusions of those different bits of reasoning by the probabilities that one attaches to p and to those alternative incompatible propositions.

    An interest-relative position with respect to the state of being “sure” is of course entirely compatible with a contextualist conception of what it is for a belief to be “justified”. E.g. we could say, ‘S is justified in believing p’ is true in a context C iff S is justified in having a high enough degree of belief in p so that, by the standards of C, justification for believing p to that degree counts as sufficient justification for being “sure” about p.

  2. Yep, this doesn’t rule out contextualism. But I think if interest-relativity is true about ‘sure’ then it’s true about knows. Interest-relativity is compatible with both contextualism and invariantism of course, which is why I left ‘invariantism’ off the property there!

    I think it’s a plausible platitude that if S knows that p, then S is sure that p, at least for the sense of ‘knows’ that philosophers care most about. As Goldman discusses, there’s a weaker sense of ‘knows’ meaning, roughly, possesses the information. But for our preferred sense, ‘knows’ requires being sure, and the best analysis of being sure is interest-relative, for pretty much the reason Mike Leigh gives.

  3. I caught an ordinary-language case (from a Ross Macdonald novel) in which ‘knows’ was clearly taken to be compatible with ‘isn’t sure’. But perhaps that’s not the sense of ‘knows’ philosophers care about. Then again, caring may be relative to the philosopher’s interests…

  4. You claim:

    A:‘knows’ requires being sure, and
    B: the best analysis of being sure is interest-relative, for pretty much the reason Mike Leigh gives.

    But even allowing A (and FWIW A’s not obvious to me) it doesn’t follow from B that ‘knows’ is interest relative does it? (Suppose e.g. there is some other condition on knowing which is only met when the strongest kind of being sure is in place …)

  5. Carrie’s point against Brian seems right; indeed it is one I emphasize about justification in my book. Even if “justified” is context-dependent, and knowledge requires justification, it doesn’t follow that “know” is context-sensitive, because maybe knowledge requires the highest kind of justification. It seems to me that she’s right as against Brian that it carries over to any argument from the interest-relativity of surity to the alleged interest-relativity of knowledge.

    On the other hand, you can replace “sure” by “know” mutatis mutandis in Mike Leigh’s dialogue, and the discourse remains perfectly natural (indeed, it is by pointing out the naturalness of discourses like this with “know” that I generally introduce IRI about knowledge to the unsuspecting philosopher).

  6. Definitions, definitions…

    If your definition of “know” is “an agent having a correct belief about the external world, then an agent may know a great many things, but no rational agent can ever know that this is true knowledge and not just strong belief, as Cartesian doubting demonstrates – and thus no rational agent can ever know. He can only believe with varying degrees of confidence.

    If your definition of “know” is commensurate with the implicit linguistic notion – if you use a definition that matches with the word’s everyday usage – then “know” would generally mean “having a high degree of confidence”, but more precisely would mean “having a degree of confidence that is relatively high given the situation”, or something similar.

    The former isn’t a very practical definition, since – using that definition – true knowledge may be extremely difficult to achieve and is generally impossible to identify. And the latter definition is clearly interest-relative.

    I think the main question is, do these definitions capture what we intend to say about knowledge? Carrie’s point seems to implicitly adopt the definition that to “know” is to have the highest possible level of confidence in a proposition. “Highest possible” seems as if it would have to be contextual if you tried to express it in terms of probabilities, unless it was simply “True” as opposed to “0.9”. In the former case, it seems there might be problems with practicality, since humans almost never assign a value of “True” to a proposition. Mostly when we say we know something, we mean we think the probability is very high, but not absolutely true. In the latter case, it seems there may still be some contextualism creeping up. Certainly at a given instant someone would either know something or not – there would only be one strongest belief appropriate to the situation, and you’d either have that level of belief or not. In the instant, contextualism is removed. But after the instant, you may be able to achieve a stronger level of belief in the same proposition, and might retrospectively qualify your earlier standpoint not as knowledge, but as mere belief. The contextualism would be exclusively temporal, but would be contextualism nonetheless.

Comments are closed.