Wesley R. Elsberry
Posts: 4991 Joined: May 2002
|
On the rarity of calculations using Dembski's EF/DI
From t.o. ...
In article <3df60bc3-robomod@ediacara.org>, Mike Goodrich <goodrich_ms@yahoo.com> wrote: >Mark VandeWettering wrote: >> In article <3df3c928-robomod@ediacara.org>, Mike Goodrich wrote:
MG> I sure hope that even though you don't know all my intentions, and even MG> though a certain amount of contrivance regarding matter, energy, and the MG> laws of physics are involved you are still making the judgment that my MG> posts are designed, thus confirming the utility of Demski 'sThree Part MG> Filter.
MV> Your posts have actually very few signs of design. But I am fascinated: MV> can you describe how Dembski's three part filter can be used to determine MV> that your postings are the result of intelligent design?
MG>Actually I think it would be far more instructive for you to describe MG>how Dembski's filter would not be useful in determing that MG>intelligent design was not the best mode of explanation for my MG>postings, asuuming that is what you think.. It would be a good MG>excercise in thinking out of the box for you. (But I won't hold my MG>breath)
I don't know that it is more "instructive", since those making the positive claim have the burden of proof. Mike's claim that Dembski's EF/DI has "utility" is a positive claim, and thus it is Mike who has the burden of proof here.
Does Mike take up his burden? Rather predictably, Mike attempts to shift the burden to others. This is simple abandonment of the claim. Mike apparently has no clue how to actually use Dembski's EF/DI, and rather than forthrightly admit this, Mike tries to distract others from recognizing this.
But Mike is not the only person for whom Dembski's EF/DI is simply too cumbersome to apply to real-world problems. Dembksi himself has attempted only four applications of varying degrees of completeness in the period from 1996 to the present. Which reminds me of the following:
[Quote]
Thus far Gell-Mann's theory has resisted detailed applications to real-world problems.
[End Quote - WA Dembski, "No Free Lunch", 2002, p.133]
Dembski's criticism of Gell-Mann's "effective complexity" is far more apposite when applied to his own concept of "specified complexity". No one but Dembski has, to my knowledge, even attempted a calculation of the sort required by Dembski's description of his EF/DI. Hmm... Actually, I may be the only other person than Dembski to attempt a calculation following his EF/DI as it was described in "The Design Inference". I seem to recall a post here in t.o. some years back showing that solutions of the "travelling salesman problem" were examples of specified complexity.
So what would have to happen for Mike to become the very first person other than William Dembski and Dembski's critics to actually apply Dembski's EF/DI, and not simply assert that it is applicable?
Dembski lays out his "argument schema" for his somewhat revised EF/DI in "No Free Lunch" on pages 72-73. Mike should refer to it for the full specification of what has to happen for an analysis to match the technical requirements of the EF/DI. Failure to fully apply this framework is rampant, as analysis of Dembski's four examples shows.
First, observe an event. It is interesting that while Dembski says that "subject S learns that an event E has occurred", Dembski is fond of using hypotheticals instead of real-world events.
Second, generate a set {H} of chance hypotheses relevant to the production of event E. This seems to be a stumbling block, for one can note that failure is common in this regard. Fully 25% of Dembski's proffered calculations (one of them) is notable for *not* including natural selection among relevant chance hypotheses (see section 5.10 of "No Free Lunch").
Third, identify a "rejection function f" and "rejection region R" such that E is in R and R "is an extremal set of f". Even Dembski skipped this part in section 5.10 of "No Free Lunch". Don't forget the gammas and deltas discussed on p.72! This requires math, not handwaving.
Fourth, identify the "background knowledge K" that "explicitly and univocally identifies the rejection function f" from step (3). Again, this step is notable by how seldom it is actually deployed, as can be seen by its absence from the discussion in section 5.10 of "No Free Lunch".
Fifth, identify the "probabilistic resources" for E "to occur and be specified relative to S's context of inquiry". BTW, Mike, S is you in this discussion. And again, even Dembski omits this step from section 5.10 of "No Free Lunch".
Sixth, fix a significance level alpha so that events less probable than alpha remains improbable conditioned on each of the chance hypotheses in {H} even when the probabilistic resources of (5) are applied. This one requires some knowledge of probability and statistics, and thus may prove more difficult for Mike than it was for Dembski.
Seventh, confirm that the probability of the rejection region R is less than alpha for all of the chance hypotheses in {H}. Again, this requires actual math, not vague handwaving, and may prove somewhat difficult for Mike.
Step 8 is just a conclusion that E exhibits specified complexity. Mike has shown no problem in jumping to conclusions regardless of the lack of warrant for them, so assuming he makes it through the preceding steps, this one should pose no difficulty. In fact, this step is so easy that most of the "examples" cited by Dembski are composed entirely of the assertion that some phenomenon E exhibits specified complexity with no accompanying justification of any sort whatsoever. In the overwhelming majority of cases, no "calculation" of any kind is offered.
If Dembski's EF/DI did have "utility" for some applications, it seems to me that someone somewhere in the six years that it has been available publicly should have picked it up and applied it to accomplish something non-trivial. Even if Mike successfully deployed the full EF/DI apparatus (an event that itself discourages breathholding), the end result (a conclusion that Mike's posts show "design" sensu Dembski) is trivial and would not support Mike's claim that Dembski's EF/DI has "utility" in any non-trivial sense.
There are other approaches to analysis of events based on algorithmic information theory that can do the useful, utilitarian tasks that Dembski talks about in making ordinary design inferences without the many drawbacks that critics have noted in Dembski's EF/DI apparatus. Wherever someone wishes to apply the EF/DI, they very likely would be better off using an alternative analytical tool. However, the alternatives do not lead to a conclusion, either deductively or inductively, of intelligent agent causation. So far, the only "utility" that has been demonstrated for Dembski's EF/DI is based not upon its "application to real-world problems", but rather in its very existence as a tool for Christian apologetics. The various failures to completely deploy the EF/DI seem to have no effect on its effectiveness in apologetics.
-------------- "You can't teach an old dogma new tricks." - Dorothy Parker
|