Discussion: Is William Dembski's CSI argument mistaken or merely useless?

By Joe Felsenstein

At Josh Swamidass’s site Peaceful Science, there has arisen some technical discussion of the detection of Design using Complex Specified Information and Alorithmic Specified Complexity. The discussion was originally about the SARS-CoV-2 virus, and how examining its sequences could shed light on whether it was designed or evolved. There was a sarcastic call for ID supporters to show that their methodology could shed light on this.

In the end, Dan Eastwoood, a biostatistician, and I fell to discussing where the fatal flaw was in William Dembski’s Complex Specified Information (CSI) criterion, and another version, his Specified Complexity criterion. We didn’t entirely agree. The discussion was technical, and did not involve theology, so I proposed moving it to Panda’s Thumb. Eastwood was willing to do this.

Below, I want to outline some issues, take a position or two, and also link to some resources for this discussion. I trust that Eastwood will join in, and others are welcome too, provided they stay on topic.

Here are some resources:

  • My article of 2007 in Reports of the National Center for Science Education on why William Dembski’s arguments don’t work, in particular his 2002 use of the Law of Conservation of Complex Specified Information
  • William Dembski’s 2005 article “Specification: The pattern that signifies intelligence”
  • A preprint version of Wesley Elsberry and Jeffrey Shallit’s 2011 article in Synthese on the difficulties with William Dembski’s CSI argument.
  • My post at PT in 2013 which explains why the conditions Dembski put on his use of the SC criterion in his 2005 article made the SC criterion a waste of time – to use it one must first answer the very question that SC is supposed to enable you to answer.
  • Dan Eastwood;’s 2016 post at his own blog explaining the problems with Dembski’s Specified Complexity quantity.
  • Dan Eastwood’s comments (for example, here and here) and my comments (here and here), and others by both of us in that thread.
  • My post here at PT last year as to why the Algorithmic Specified Complexity (ASC) measure is irrelevant to any argument about barriers to adaptation in evolution.

Let me start off the PT discussion by offering some possibly-controversial opinions:

  1. Contrary to what a lot of my fellow defenders of evolutionary biology declare, Complex Specified Information is not a meaningless concept. It is just that seeing it does not constitute evidence for Design.
  2. Algorithmic Specified Complexity (ASC) is not a sensible quantity to compute to evaluate how easy or hard it is to evolve adaptations. In fact it has nothing to do with adaptations and is best ignored.

So, let’s discuss. I will, as usual pa-troll this discussion against off-topic crap.