A peer-reviewed article that supports ID . . . or something else

By Joe Felsenstein, http://www.gs.washington.edu/faculty/felsenstein.htm

William Dembski and Robert Marks have published what Dembski describes as a “peer-reviewed pro-ID article”. It is in the computer engineering journal IEEE Transactions on Systems, Man and Cybernetics A, Systems & Humans in the September 2009 issue. In a post at his Uncommon Descent blog (where a link to a PDF of the article will also be found) Dembski describes it as critiquing Richard Dawkins’s “Methinks it is like a weasel” simulation and that “in critiquing his example and arguing that information is not created by unguided evolutionary processes, we are indeed making an argument that supports ID.” But what does it really say about ID?

The article does not mention ID directly, but defines a quantity called “active information” in search problems. Basically, it measures how much faster the solution can be found by a search in a problem’s space than by looking for the solution by drawing points from the space in a random order — how much faster one finds the solution than a monkey with a typewriter would. In Dawkins’s Weasel case, a monkey with a typewriter finds the solution after about 1040 tries, while one version of Dawkins’s program would take only about 728 tries. The active information is the log of the ratio of these numbers, about 124 bits.

In effect, the picture the article paints is that information is out there in the shape of the fitness surface — the way fitnesses change as we move among neighboring genotypes. So, on this view, natural selection does not create information, it just transfers it into the genotype. The information is out there already, lying around. Dembski and Marks at one point say that “the active information comes from knowledge of the fitness”. If the fitness surface is smooth, as in the Weasel case, natural selection will readily be successful. D&M would then regard the information as provided by a Designer in advance.

In that case natural selection works. If a Designer has structured our genotype-phenotype space so that fitness surfaces are often smooth, if mutations do not typically instantly reduce the organism to a chaotic organic soup, if successful genotypes are often found to be close in sequence to other successful genotypes, then the Designer is not designing individual organisms — she is leaving natural selection to do the job. Dembski and Marks’s argument would then at most favor theistic evolution and could not be used to favor ID over that.

One can wonder whether one needs any particular Designer to structure reality in that way. The laws of physics do not make all objects interact intimately and strongly. When I move a pebble in my back yard, the dirt, grass, trees, and fences do not instantly reorder themselves into a totally different arrangement, unrecognizably different. If they did, of course natural selection would not be able to cope. But as they interact much less strongly than that, only a few leaves of grass change noticeably. I can cope, and so can natural selection. Does the smoothness of fitness surfaces come from this weakness of long-range interactions in physics? If so, then Dembski and Marks’s argument ends up leaving us to argue about where the laws of physics ultimately come from, and most evolutionary biologists will not feel too worried.