What timing! Dembski again demonstrates innumeracy

Jun 16 2006 Published by under Debunking Creationism

Right after finishing my post about how Dembski has convinced me that he is not a competent mathematician, I find PZ linking to a Panda's Thumb post about Dembski, which shows how he does not understand the meaning of the mathematical term "normalization".
Go look at the PT post: Something rotten in Denmark?
Is this guy really the best mathematician the ID folks have available to represent them?

26 responses so far

  • bigdumbchimp says:

    Is this guy really the best mathematician the ID folks have available to represent them?
    He's also the best biologist they have to represent them.

  • Uffe says:

    Being from Denmark, this triggered my curiosity to the point of sampling some of the Danish sites Google finds in the search. That showed the whole thing to be even more rotten above ignorance of normalization: The Danish sites don't actually talk much about ID itself, instead I read articles discussing the perceived decline of science in US, the risk of US becoming a theocratic state, etc. In other words ID as a symptom, not the illness itself. To count all this as ID support makes Dembski a villain in addition to a poor mathematician.
    Remember the recent story about the cartoons from Denmark that made a crisis in the Middle East? Well, in all that I also found a cartoon about ID in one of the major newspapers. You don't need to understand Danish to get the point of the article:
    Have a look (Dembski counted this one too!):
    ... but, please, don't have ID supporters burn down Danish embassies this time.

  • You are not qualified to judge, especially given your inarticulate, ill-informed criticism of the first step of Richard Swinburne's Bayesian argument. Assigning a prior probability of .5 to an event and its complement, respectively, when the probability of the event of interest is unknown is a conservative approach. (Of course, that prior should be updated with relevant information.)

  • Robert:
    You're welcome to your own opinion; but I think that Swinburne's claim that he can "prove" the correctness of Christianity by just blindly asserting 50/50 probability for "God exists" and "God would choose to be incarnated" is, to put it mildly, bullshit.
    I'm not a fan of the bayesian approach of saying that you can always start by just asserting 50/50 odds. I understand the idea behind it, which is that if you want to work out what the probability of something is, you need a starting point, and 50% is a workable null hypothesis. But I prefer the approach of saying that if you have absolutely no data about the actual probability of an event, then you can't assign it a probability.
    I find the Bayesian 50/50 null hypothesis leads to a lot of sloppiness like Swinburne's nonsense. He pulls the 50/50 nonsense out of his hat, and claims that he can use it not because it's actually a valid position, but because it's the initial null hypothesis, and he doesn't want to pay any more attention to working out a meaningful probability analysis.
    Using the unmodified null hypothesis lets you create all sorts of nonsense arguments: just take a bunch of unrelated ideas, assign each of them the 50/50 null hypothesis probability, and then do the bayesian combination.
    Want to make the probability of the truth of Christianity 90%? No problem. Assemble things the way Swinburne did. Want to make the probability of the truth of Christianity 10%? No problem. Just pick a different set of pseudo-events, don't waste time actually reasoning about how probable any of them are, and just give 'em each the null hypothesis. Presto! You get the result you wanted.
    If you want to do Bayesian probability, you need to be careful to make sure than your events are independent when you assert that they're independent; that they're atomic when you assert that they're atomic; and that you actually use all of the knowledge available to you to to assess the probability of each event instead of just inserting null hypotheses whenever it's easier.
    Swinburne didn't do any of that. He asserted the independence of non-independent things; he used compounds instead of atoms; he use null hypotheses whenever he didn't feel like actually reasoning about the real probabilities of events; and he just pulled probabilities for other events out of thin air for no reason.
    For you Bayesians out there: I know that Swinburne's way of pulling the null hypothesis out whenever he's feeling lazy is not a valid use of Bayesian statistics.

  • Mark:
    I did not claim all of Swinburne's steps were justified, just the first. (As I recall, that is the only one you specifically addressed.)

  • Robert:
    Actually, you're wrong. I specifically talked about the first two steps of his gibberish and the way he connected them; I thought was more that adequate for something as pathetically stupid as that.
    I quite seriously think that the "God exists" = 50/50 and "God incarnates" = 50/50 is no more reasonable than "Flying pink monkeys fly out of my butt"=50/50, and "flying pink monkeys attack his house"=50/50. The fact is, both are just stupid and ridiculous. The only reason that Swinburne's argument has any more claim to respectability than my flying pink monkeys argument is because we cut slack to any argument that involves religion, especially christianity.

  • Mark:
    I am not interested in defending anything beyond Swinburne assigning God's existence a prior probability of .5, which is perfectly legitimate. Of course, leaving it like that, without updating the prior, is not particularly informative.

  • Ross says:

    Am I missing something, or is there not also something weird about Dembski saying "international interest in ID is growing" because he thinks that a higher percentage of people in other countries search for it?
    I think he might have also forgotten his basic calculus if he thinks that the value of a statistic is the same as its rate of change...

  • Robert:
    What is your point?
    As I said in the original post, and as I've said here: the point is that using the 50/50 null hypothesis in that argument is pure, utter bullshit. The 50/50 null is fine as a starting point if you're looking at an atomic event in order to compute a meaningful probability for it. "God exists" in Swinburne's argument is *not* an atomic event. It's a compound statement: "*A* god exists" and "Only one god exists" and "The god that exists is the Christian god" and ...
    Assigning that a 50/50 probability is bullshit.

  • Torbjörn Larsson says:

    "I am not interested in defending anything beyond Swinburne assigning God's existence a prior probability of .5, which is perfectly legitimate. Of course, leaving it like that, without updating the prior, is not particularly informative."
    Not considering Marks' argument, since Swinburne can't update the prior he doesn't raise the it above the default assumption. But the largest question for me is if Bayesian probability assumes that you can improve your apriori belief. If it does, and Swinburne can't make observations, it is a failed model.

  • Mark:
    The God (http://www.iep.utm.edu/g/god-west.htm) Richard Swinburne has in mind precludes the possibility of other gods existing. My answer to "But what about other gods?" is that they are lumped into the complement.
    BTW, are you a fellow Irishman?

  • Torbj#$%N: 😉
    I disagree that Swinburne can't update his prior. He could update it by making recourse to standard arguments for God's existence.

  • Robert:
    Swinburne is welcome to his own belief of what form a deity/deities take in his belief. But if he wants to make a probabilistic argument, and treat the statement of the existence of *his* god as an event, then he can't arbitrarily exclude other possibilities, and pretend that the existence of his god is an atomic statement, whereas any consideration of other religion's deities are necessarily not atomic. (Which is what is implied by "Probability his god exists = .5, sum of probabilities of all other propositions concerning existence or non-existence of other deities =.5")
    And no, I'm not an Irishman. I know it's confusing, given that "Carroll" is Irish-sounding, and that I both listen to and play traditional Irish music; but the "Carroll" part of my name is actually an Americanization of the Galitzian Jewish name "Karolciok" (pronounced, roughly, "Cuh ROLL chuck"). The "Chu" part of my name comes from my wife; we combined our names when we got married; her part came first because when we got married, she'd published a dozen papers as "Jennifer Chu", and I'd only published one paper as "Mark Carroll". So putting the "Chu" first would help with the paper trail for people following her research.
    So I'm an Ashkenazi Jewish guy with a half-Chinese, half-Irish sounding name who plays Irish music.

  • Canuckistani says:

    On the topic of justification for the Principle of Indifference as a general principle for assigning probabilities, I recommend the argument expounded in the middle of the second chapter of this document. The upshot is that the principle can be derived from a more primitive axiom, namely that equivalent states of information yield equivalent probabilities.
    In terms of God, the 50%/50% probability assignment is only justified if one's information is symmetric about the existence/non-existence of God. This might be justified for the sort of God who touches off the Big Bang and then sits back, but for specifically one dude's concept of the Christian God? Not so much.

  • Canuckistani:
    Thanks, that's what I was trying to say, but you phrased it much more clearly than I did.

  • Andrew Wade says:

    The God (http://www.iep.utm.edu/g/god-west.htm) Richard Swinburne has in mind precludes the possibility of other gods existing. My answer to "But what about other gods?" is that they are lumped into the complement.

    That made me think of the following student blooper, which seemed a propos:

    "Handel was half German, half Italian, and half English. He was very large."

  • You ask: "Is this guy really the best mathematician the ID folks have available to represent them?" You might like the following passage from the same mathematician:
    "Despite my disagreements with Morris and young earth creationism, I regard those disagreements as far less serious than my disagreements with the Darwinian materialists. If you will, young earth creationism is at worst off by a few orders of magnitude in misestimating the age of the earth. On the other hand, Darwinism, in ascribing powers of intelligence to blind material forces, is off by infinite orders of magnitude." (In: "Intelligent design's contrubition to the debate over evolution: a reply to Henry Morris", 2005)

  • Ricardo:
    He actually said that? "Infinite orders of magnitude"?
    Ye Bloody Gods, what an idiot...

  • Yes he did. He actually wrote it. (The link isn't working right now for some reason, but it should come back.)

  • Torbjörn Larsson says:

    Updating beliefs with more beliefs instead of observations doesn't seem to be what Bayesian inference is about, but I can see that one could possibly accumulate beliefs.
    The problem for Swinburne then becomes that we are allowed to input all our beliefs that affects his Bayesian inference in the other direction. Mark already complained about his arbitrary exclusion rule.

  • Mark:
    The uninformative (or objective) prior which assigns equal probability to each event maximizes the entropy. Since there are only two (mutually exclusive) events here (i.e., God exists or He does not), Swinburne's choice was justified.

  • Robert:
    No, there are not only two mutually exclusive events here. That's one of Swinburne's biggest errors.
    Swinburne's claim is that there are only two possibilities: (1) Swinburne's Christian god exists, or (2) no god(s) at all exist.
    But (1) is not an atomic event. It's multiple statements mushed together for no reason in order to stack the probabilities in his favor. You don't get to start from complex compound statements without separating the distinct components and analyzing them separately.
    It is, in essense, setting things up so that there's a 50% chance that *my religion is true*, but the sum of the probabilities of all other beliefs in deities, as well as the probability of their being no deities at all, are all considerably less that 50%. Why does Swinburne's concept of deity get a higher probability? Because he wants it to. He has no more justification than that.

  • Bronze Dog says:

    Either I am on Mars or I am not.

  • PaulC says:

    Bronze Dog:

    Either I am on Mars or I am not.

    Earth would work just as well, but:
    If you leap into the air from a standing position on Mars, most people would probably agree you're still on Mars. If you're in a rocket launched from Mars even in a suborbital trajectory, most people would say that at some point you're not on Mars, though I doubt they'd agree on where. You can define it rigorously to make it unambiguous, but I think the informal statement has a lot of middle ground. I have only the slightest familiarity with fuzzy logic, but isn't this kind of thing the motivation behind it?

  • Bronze:
    I think a better example of what's wrong with the first point of Swinburne's argument is:
    There's a 50% chance that I'm a french citizen. Either I'm french, or I'm not.
    The problem with statement is pretty much exactly the same as the problem with "Either God exists, or he doesn't".

  • zilch says:

    I only know what I remember from high school about statistics, but I can assure you that there's a 50% chance that the Invisible Pink Unicorn is eating a piece of invisible Austrian cheese from my hand right now. God doesn't fit this analogy? Too big, or too omnimax? Lol.

Leave a Reply