Wesley R. Elsberry
Posts: 4991
Joined: May 2002
|
Jeff Shallit and I have a topic on replacing Dembski's notion of specification in the appendix of our long paper critiquing William Dembski's ideas . (See Information Theory, Evolutionary Computation, and Dembski's "Complex Specified Information".)
In the article, we note a number of problems with "specification" as given by Dembski. Our replacement, specified anti-information (SAI), does not have the drawbacks that we note for specification. SAI is grounded in algorithmic information theory and can be considered an application of the universal distribution. (See the original article by Kirchherr, Li, and Vitanyi.)
Some of our colleagues reading the drafts of our long paper back in 2002 were non-plussed: why did we attempt to "rescue" the notion of specification with a replacement? The answer is that we felt that some positive statement needed to be made rather than making an entirely negative critique. Dembski's examples resonate with readers, so we felt that a non-Dembskian approach was needed to show that the examples could be dealt with in a rigorous way, but that the further conclusions about intelligent agency that Dembski urges were unsupported.
SAI accomplishes both these goals. SAI is easy to apply to problems that can be reduced to a bit-string representation, as Dembski has done for some of his examples. SAI also warrants a far weaker implication than intelligent agency as a cause: an event with high SAI is likely caused by a simple computational process. Elsewhere in the article, we discuss the ubiquity of natural computational processes. There is no "design inference" that can be based upon SAI, just a "simple computational process inference".
Edited by Wesley R. Elsberry on Nov. 24 2003,10:21
--------------
"You can't teach an old dogma new tricks." - Dorothy Parker
|
