Dembski and "No Free Lunch", Reprise

By Wesley R. Elsberry
"Intelligent Design" advocate William A. Dembski recently announced his completion of another draft paper. Titled "Searching Large Spaces: Displacement and the No Free Lunch Regress", Dembski's paper is an assertion, with math, that he was right about "No Free Lunch" all along. An interesting point is that Dembski announced this paper to various critics via an email with the paper attached. I was on this list of email recipients, and am proud to share that place with Eugenie Scott, Richard Dawkins, Daniel Dennett, and fourteen other critics. Another of the critics, David Wolpert, was a co-discoverer of the "No Free Lunch" theorems. Since I was on the original mailing list, I've been treated to an exchange of emails between Wolpert and Dembski. Wolpert pointed out to Dembski that there were, in fact, "free lunches" in coevolution, and attached a copy of a paper by Wolpert and MacReady on that topic currently in submission. Dembski responded that Dembski's result "subsumed" Wolpert's findings. Wolpert responded to that, asking that Dembski actually read his paper before criticizing it. Dembski responded again, re-asserting his claim and adding the conjecture that Wolpert's status as co-discoverer of "No Free Lunch" might actually be preventing Wolpert from thinking clearly about these issues. This is shaping up as something worthy of a big bag of popcorn, except that my doctor has prohibited that in my case. Another interesting point is how Dembski has gone from trying to appear like he has been responsive to critics to once again giving his critics the cold shoulder. Despite the various commentaries and critiques of Dembski's book No Free Lunch one will discover that only one criticism is acknowledged and referenced in Dembski's new paper. The sole critic to get a mention was Michael Shermer. Here is the passage:

It follows that assisted search, even with so modest a problem as finding a specific protein 100 amino acids in length, requires a considerable amount of information if it is to surpass blind search and successfully locate a target. How are we to explain this net increase in information? One way is to explain it away by suggesting that no targets are in fact being searched. Rather, a space of possibilities is merely being explored, and we, as pattern-seeking animals, are merely imposing patterns, and therefore targets, after the fact (see, for instance, Shermer 2003).

This explanation may work in certain instances where humans make up patterns as they go along. But many patterns-whether in the arts or in engineering or in the natural sciences-are objectively given. For instance, it is an objective fact whether a given polymer has a certain strength and resilience. Thus, searching through a polymer configuration space to find a polymer with at least that level of strength and resilience constitutes a search for an objectively given pattern qua target. If such a polymer is found and if the target within which it resides has small uniform probability, then a considerable amount of information needs to be incorporated in an assisted search for it to be successful, a fact that will be reflected in the information measure I as applied to the canonical exchange probability (i.e., I(ยต0|U)).

It certainly doesn't look like Dembski is "answering the toughest questions about intelligent design" here. It looks much more like Mark Perakh is right that Dembski is evading the toughest questions (and, in fact, N-1 of the questions) concerning his claims about NFL, regress, and displacement. And here's a particularly silly piece of prose by Dembski:

Intelligence acts by changing probabilities. Equivalently, intelligence acts by generating information. For instance, a slab of marble temporarily has a high probability of remaining unchanged. Then, without warning, Michelangelo decides to sculpt David, and the probability of that marble slab taking on a new form (i.e., receiving new information) now changes dramatically.

Michaelango did not "act by changing probabilities". Michaelangelo acted by taking a hammer and chisel and whacking off the pieces of marble that weren't David. The statue of David was sculpted by the expedient of physical action, not by "changing probabilities". Michaelangelo could have cogitated all he wanted concerning how he might like the piece of marble to look when he was done without having the slightest effect on the shape of the marble right up until the time he actually applied the chisel to the rock. The "probabilities" are a topic of our analysis, and not a mechanism of action. Dembski has long had a fascination for the assertion that intelligence "generates" information. Yet Dembski has been remarkably reticent about saying how this happens. The whole "apparent CSI" vs. "actual CSI" split of Dembski's turns on Dembski's failure to address the issue of whether information is "generated" when an effective method is applied to a problem. Is information "generated" when an effective method is applied, or only when an effective method first arises? Dembski has never taken up this question and given a satisfactory answer. Part of his reticence may be that if mere application of an effective method does not qualify as "generating" CSI, then the only way left for "generating" CSI by intelligent agents is via irrationality. It is unlikely that Dembski wants to assert the creation of CSI by his preferred intelligent designer comes from irrational processes.