Friday, July 6, 2007

The Explanatory Filter Deconstructed

In the comments on the first post, c.loach points us to a fascinating discussion of the Explanatory Filter. As loach writes:
Watch as scordova worms his way out of doing a demonstration of the The Explanatory Filter, the masterwork of ID that does not work!
Much better than the EF itself is Håkan Rosén's brilliant Imploratory Filter:

1. Does the text advocate a pro-ID stance?
2. Is it possible than someone could have written the text as a joke?
3. Does it look as if the writer is trying to hide the fact the the writing is a hoax?

Only if the answers to the first two questions are yes do we proceed to question three. This is important since we know that people do write pro-ID hoaxes. Pro-ID hoaxes just don't materialize from writings about Goethe or Homer. The third stage of the Imploratory Filter presents us with a binary choice: attribute the thing we are trying to examine to deliberate deception if it appears joke-like; otherwise, attribute it to self-deception. In the first case, the writing we are trying to examine is not only pro-ID, but also appears joke-like. In the other, it is pro-ID, but appears deluded. It is the category of joke-like writings having a pro-ID stance that reliably signals a hoax. "Non-funny" writings advocating ID, on the other hand, are properly attributed to self-deception.

The last thing we need to consider is the case of false positives and false negatives. This method can, unfortunately, yield false negatives. It is possible that some piece of writing might be labelled a non-hoax, when it in fact is a hoax. On the other hand, the method yields no false positives. I.e., when the filter claims that a writing really is a pro-ID hoax, it will will never turn out to be a non-hoax.

The Imploratory Filter faithfully represents our ordinary practice of sorting through things we alternately attribute to self-deception or hoaxes. In particular, the Filter describes:
* how Michael Egnor is still allowed to post for the DI.
* how Casey Luskin can keep repeating that ID can make predictions.
* how Dembski can claim that the explanatory filter yields no false positives even though it measures design via specified complexity of which irreducible complexity is a subclass. Irreducible complexity, in turn, allows for false positives.

Apologies for reposting this, but the IF may well shake the foundations of Dembskism.


Thursday, July 5, 2007