Bayesian constraint by the evidence (was: Why believe the truth?)

From: Dan Fabulich (dfabulich@warpmail.net)
Date: Wed Jun 18 2003 - 17:39:29 MDT

  • Next message: Robin Hanson: "RE: The Future of Secrecy"

    Eliezer S. Yudkowsky wrote:

    > Dan Fabulich wrote:
    > >>
    > >> Nature doesn't work that way in constructing explanations; how could it
    > >> be a good method for discovering them?
    > >
    > > My first response: "I would least expect to get this kind of argument
    > > HERE, of all places! Isn't it rather the point that we can do a bit
    > > better than nature?"
    >
    > When I am *designing* something, then yes, I will try and make the design
    > modular because that is a good heuristic for humans to use. When I am
    > trying to *discover* something I will not try to make the *explanation*
    > modular unless I think the *reality* is modular - the purpose of the map
    > is to correspond to the territory, after all. If you can build a theory
    > that's modular, just because you want it to be modular, doesn't that mean
    > you're inventing something rather than discovering something? Is this not
    > the very essence of "mere philosophization", building maps unconstrained
    > by territory?

    I think you're getting yourself confused with the language of
    "unconstrained by territory". I also note that you simply snipped my
    Bayesian argument, which I took to be the meat of my point.

    Given a theory T and our background theory B, we regard our theory T to
    *more* likely if P(T|~B) is higher, *regardless* of P(B), all else being
    equal. Do you contest this for some reason?

    Let's suppose I agree with you that any theory T for which P(T|B) and
    P(T|~B) are both high is less "constrained by the territory" than an
    alternate theory T', under which P(T'|B) = P(T|B), but P(T'|~B) is very
    low. T' is constrained by the territory. T is less constrained by the
    territory. Are you trying to tell me, against Bayes, that we should hold
    T' to be more likely than T, because T is "mere philosophy" whereas T' is
    "constrained by the facts"?

    Now, don't get me wrong: being unconstrained by the territory CAN be a bad
    thing. But it's a bad thing only if you are philosophizing *instead* of
    using all the data that's available.

    (This can happen in at least two seemingly different ways: you may either
    make a claim that is too weak, because you refrain from making a deeper
    correct judgment when only allowing yourself access to some subset of the
    data, or you may make a claim that is false [or not supported as strongly
    as you might think], because it would be contradicted by further evidence
    that you refuse to consider. A Bayesian would assimilate both types of
    errors as mistakes about the likelihood of certain propositions, either
    too low or too high.)

    But that's not at all the case in my T & T' example, by construction, and
    it doesn't apply to our "Why believe the truth" argument either.
    Nobody's saying that we shouldn't consider the probability of a
    Singularity [P(T|B)] in our calculations. But we ARE saying that a theory
    that applies well to Singularitarians and non-Singularitarians alike is a
    better theory, more likely to be true, than a theory that applies equally
    well to Singularitarians but not at all to non-Singularitarians, ceteris
    paribus.

    Indeed, this very argument, is all-the-stronger because it has nothing to
    do with being an academic; it has everything to do with the consequences
    of accepting Bayes.

    -Dan

          -unless you love someone-
        -nothing else makes any sense-
               e.e. cummings



    This archive was generated by hypermail 2.1.5 : Wed Jun 18 2003 - 17:49:51 MDT