Wesley R. Elsberry
Posts: 4991
Joined: May 2002
|
Learned Hand,
I'd be interested in your thoughts on how Dembski has changed his arguments since 2001. If you go to the Useful Links page, there are links to video there from the Haverford "Interpreting Evolution" conference in 2001. Dembski presented, and I gave a critical presentation. You might check out the videos where I appear or have a look at my PowerPoint, which is also linked from that page.
In 2001, I pointed out that in order to test his notions, Dembski would need to concentrate on examples that already have good natural explanations, and see whether whether his technique indicates "design" if you exclude that explanation. I specifically mentioned the Krebs citric acid and mammalian middle-ear impedance-matching anatomy then.
In 2002, John Wilkins and I published a critique of Dembski's "explanatory filter" that laid out in greater detail our general objection to treating "ordinary design inferences" as being of the same class as "rarefied design inferences".
In 2003, Jeff Shallit and I posted a lengthy critique of Dembski's "complex specified information". It contains a section on possible test cases for application of CSI. It also contains an appendix on "Specified Anti-Information" (SAI) that applies the universal distribution to the idea of testing to exclude chance as the source of information. The universal distribution is entirely based on algorithmic information theory and owes nothing to Dembski's probabilistic and logic-chopping approaches. Here's something I said about SAI outside of that essay:
| Quote |
The existence of a minimal program/input pair that results in a certain output indicates that there exists an effective method for production of the output. Since effective methods are something that are in common between intelligent agents and instances of natural computation, one cannot distinguish which of the two sorts of causation might have resulted in the output, but one can reject chance causation for the output. We haven't so much repaired specification as we have pointed out a better alternative to it.
This leads me to a claim about Dembski's design inference: Everything which is supposedly explained by a design inference is better and more simply explained by Specified Anti-Information.
SAI identifies an effective method for the production of the output of interest. The result of a design inference is less specific, being simply the negation of currently known (and considered) regularity and chance. The further arguments Dembski gives to go from a design inference to intelligent agency are flawed. On both practical and theoretical grounds, SAI is a superior methodology to that of the design inference.
|
The section on challenges for IDC advocates was published in Reports of the NCSE in 2008 (available online).
Dembski certainly knows about these critiques. It doesn't sound like he acknowledged them within his presentation.
--------------
"You can't teach an old dogma new tricks." - Dorothy Parker
|
