Re: Examples of natural selection generating CSI

From: Richard Wein (rwein@lineone.net)
Date: Sat Nov 04 2000 - 09:09:27 EST

  • Next message: silk: "Nothing comes from nothing or does it?"

    From: Ivar Ylvisaker <ylvisaki@erols.com>

    >Richard Wein wrote:
    >
    >> From: Ivar Ylvisaker <ylvisaki@erols.com>
    >>
    >> >Second, Dembski specifically disavows that this is what he means by a
    >> >design inference. He writes a little later in the reference cited
    >> >above:
    >> >
    >> >"Thus, we shall never see a design hypothesis D pitted against a
    >> >chance hypothesis H so that E confirms D better than H just in case
    >> >P(D|E) is greater than P(H|E). This may constitute a "Bayesian
    >> >design inference," but it is not the design inference stemming from
    >> >the Explanatory Filter."
    >>
    >> >He seems to waver a bit about this conviction in his "Intelligent
    >> >Design" though the book still contains a description of the filter.
    >> >See "abduction" in the index.
    >>
    >> I haven't got "Intelligent Design". Perhaps you could quote me the
    relevant
    >> passage, if it's not too long.
    >
    >"Abduction" and an "inference to the best explanation" are similar
    >concepts. Finding a relevant passage using the latter term is easier.

    The term "abduction" is new to me. For anyone else who's interested, I found
    the following relevant pages:
    http://www.cis.ohio-state.edu/lair/Projects/Abduction/abduction.html
    http://www.rz.uni-frankfurt.de/~wirth/inferenc.htm
    http://carbon.cudenver.edu/~mryder/itc_data/abduction.html

    I found also the following review of the Josephsons' book "Abductive
    Inference":
    http://coombs.anu.edu.au/Depts/RSSS/Philosophy/People/Cathy/Josephsons.html

    I was interested to see the reviewer make the following comment, which
    relates to the arguments I've been making in my previous posts:

    "Also noteworthy in this formulation is the LAIR researchers' emphasis on
    the importance of ruling out alternative explanations as a means of arriving
    at positive confidence when reasoning abductively."

    I was also interested to read:

    "The book closes with a discussion of the concept of "plausibility" (of
    hypotheses) which the authors lean heavily on throughout the book. They
    explore the relationship of plausibility to the traditional formally
    elaborated "probability", and also the features which make this concept
    unique, such as the strong role which analogy may play in the generation of
    plausible hypotheses."

    >On page 274 of Intelligent Design, Dembski quotes Sober ("Philosophy
    >of Biology," page 33):
    >
    >[Start Sober's quote; Dembski added the material in []]
    >
    >"[Paley's] argument involves comparing two different arguments -- the
    >first about a watch, the second about living things. We can represent
    >the statements involved in the watch argument as follows:
    >
    >A: The watch is intricate and well suited to the task of timekeeping.
    >W1: The watch is the product of intelligent design.
    >W2: The watch is the product of random physical processes.
    >
    >"Paley claims that P(A|W1) >> P(A|W2) [i.e., the probability of A
    >given that W1 is the case is much bigger than the probability of A
    >given that W2 is the case]. He then says that the same pattern of
    >analysis applies to the following triplet of statements:
    >
    >B: Living things are intricate and well-suited to the task of
    > surviving and reproducing.
    >L1: Living things are the product of intelligent design.
    >L2: Living things are the product of random physical processes.
    >
    >"Paley argues that if you agree with him about the watch, you also
    >should agree that P(B|L1) >> P(B|L2). Although the subject matters of
    >the two arguments are different, their logic is the same. Both are
    >inferences to the best explanation in which the Likelihood Principle
    >[a statistical principle which says that for a set of competing
    >hypotheses, the hypothesis that confers maximum probability on the
    >data is the best explanation] is used to determine which hypothesis
    >is better supported by the observations."
    >
    >[End Sober quote]

    I haven't read Paley. Do you know if Paley's argument is indeed a maximum
    likelihood one? It's not unusual for Dembski to present his own
    interpretation of a writer as if it was what that writer actually wrote.

    Anyway, this argument may have been reasonable in Paley's day, when natural
    selection was unknown. But it isn't reasonable today. P(B|L2) is high,
    because natural selection ensures that living things are "well-suited to the
    task of surviving and reproducing." On the other hand, we have no idea what
    P(B|L1) is, since L1 says nothing about the aims and capabilities of the
    designer.

    Furthermore, maximum likelihood arguments are useless when they ignore other
    relevant information (such as all the evidence that we have in support of
    natural evolution). To take an extreme example:

    O (Observation): A roulette ball lies in the "36" slot of the wheel.
    H1: The wheel was spun normally.
    H2: The croupier simply placed the ball in that slot.

    Then P(O|H2) > P(O|H1), so, by the likelihood principle, we should prefer
    H2. But, if we watched the roulette wheel carefully, and saw that the
    croupier did *not* interfere with the ball, we would be foolish to accept
    H2.

    >[Ivar: In mathematics, a vertical bar "|" is equivalent to the words
    >"given or given that," e.g., P(B|L1) means the probability of B
    >given that L1. The notation ">>" means "much greater than."]
    >
    >Dembski continues on the next page:
    >
    >"Inference to the best explanation is inherently competitive (cf.
    >section 7.4). Best explanations are not best across all times and
    >circumstances. Rather they are best relative to the hypotheses
    >currently available and the background information we have to
    >evaluate those hypotheses. Sober therefore has to leave the door open
    >to design even though he doesn't think it very likely that design
    >will ever pose a serious threat to Darwinism. He concedes, "Perhaps
    >one day, [design] will be formulated in such a way that the auxiliary
    >assumptions it adopts are independently supported. My claim is that
    >no [design theorist] has succeeded in doing this yet." [snip ref.]
    >The burden of this book has been to show that design remains a live
    >issue and can once again be formulated as the best explanation for
    >the origin and development of life."

    Thanks Ivar, that's very interesting. It shows Dembski being equivocal as
    usual.

    >The thesis of Dembski's "The Design Inference" (TDI) is not an
    >inference to the best explanation. In Sober's notation, TDI asserts
    >that if one can show that P(L2) is essentially zero, then P(L1) must
    >be very nearly one. An inference to the best explanation asserts
    >that design is a better explanation than random processes if
    >P(B|L1) >> P(B|L2).

    You seem to be suggesting that "inference to the best explanation"
    necessarily means a maximum likelihood inference. My impression is that it
    has a wider meaning. Actually, most IDers seem to just use "this is an
    inference to the best explanation" as a more impressive sounding way of
    saying "I think this is the best explanation." ;-)

    >Using the same notation, here is Bayes's rule:
    >
    >P(L1|B) = ( P(B|L1)P(L1) ) / ( P(B|L1)P(L1) + P(B|L2)P(L2) )
    >
    >where P(L1) and P(L2) are the probabilities of L1 and L2 before using
    >the observation B. Bayes's rule is mathematically correct. However,
    >Bayes's rule remains controversial because of the problems of
    >assigning values to P(L1) and P(L2).
    >
    >It is the value on the left side of Bayes's rule that people really
    >care about. However, if P(L1) is about equal to P(L2), then
    >P(B|L1) >> P(B|L2) leads to the same conclusion, i.e., the inference
    >to the best explanation points to the same answer.
    >
    >Note that if Dembski can show that P(L2) is essentially zero, then
    >Bayes's rule says that P(L1|B) is essentially one, i.e., intelligent
    >design is true.
    >
    >On the other hand, if assigning a persuasive value to P(L2) proves
    >difficult, then the inference to the best explanation is a plausible
    >fallback position. Nevertheless, it is a different and much less
    >ambitious argument.

    Isn't a Bayesian inference also an "inference to the best explanation"?

    I found the following very interesting article on the likelihood
    principle, by Forster and Sober:
    http://philosophy.wisc.edu/forster/Likelihood/WhyLikelihood.htm

    By tbe way, I note that you didn't reply to the main part of my last post,
    and just when I thought it was getting interesting. Is this because you (a)
    agreed with me, (b) lost interest, (c) thought that we'd come to an impasse,
    or (d) something else? ;-)

    Richard Wein (Tich)
    --------------------------------
    "Do the calculation. Take the numbers seriously. See if the underlying
    probabilities really are small enough to yield design."
      -- W. A. Dembski, who has never presented any calculation to back up his
    claim to have detected Intelligent Design in life.



    This archive was generated by hypermail 2b29 : Sat Nov 04 2000 - 09:08:12 EST