Weaseling It Right At Last
In chapter 19 of Being as Communion (2014), the “intelligent design” promoter William A. Dembski returns to METHINKS IT IS LIKE A WEASEL, a string of letters that owes its fame to The Blind Watchmaker (1986), by Richard Dawkins with additional dialogue by William Shakespeare. In The Blind Watchmaker, Dawkins illustrates the power of repeated bouts of variation and selection by showing how the famous phrase from Hamlet can be generated from a randomly generated sequence of twenty-eight letters and spaces by a simple algorithm. “If evolutionary progress had had to rely on single-step selection, it would never have got anywhere,” Dawkins concludes. “If, however, there was any way in which the necessary conditions for cumulative selection could have been set up by the blind forces of nature, strange and wonderful might have been the consequences” (emphasis in original)—including, as it happens, the whole splendid diversity of life.
Dembski, in his 2014 book, describes the algorithm of The Blind Watchmaker thus: “take the current sequence, sloppily copy it 100 times, with a 5 percent mutation rate (i.e., chance of error) at each character, and then select the (randomly modified) copy with the best score, making it the new sequence for the next generation.” (The “score” equals “the number of character matches with the target sequence.”) It might seem as though there’s nothing remarkable in Dembski’s description of the algorithm. And indeed there isn’t. (Dembski supplies his own numbers for the number of copies and the rate of mutation, and like Dawkins he fails to specify what happens when copies have equal scores, but these are unimportant matters.) What’s remarkable about it is that finally, after fifteen years of misdescribing the algorithm, persisting in doing so despite public correction, Dembski has finally described it correctly.
The story begins in 1999, when Dembski’s essay “Explaining Specified Complexity” was posted at the Metanexus website (which, back in the day, was a venue in which “intelligent design” promoters and their critics often tangled). In it, he described the algorithm in the following way:
(i) Start out with a randomly selected sequence of 28 capital Roman letters and spaces …; (ii) randomly alter all the letters and spaces in this initial randomly-generated sequence; (iii) whenever an alteration happens to match a corresponding letter in the target sequence, leave it and randomly alter only those remaining letters that still differ from the target sequence.
Do you see the difference? In the “Explaining Specified Complexity” version, the algorithm latches, as it were, onto the correct letters: once a correct character from METHINKS IT IS LIKE A WEASEL appears, it never disappears. In the original version presented in The Blind Watchmaker, there is no suggestion that the algorithm latches.
The same description was aired in April 2000, when Dembski gave a paper at a conference, The Nature of Nature, at Baylor University, where he was then running the Michael Polanyi Center, which Dembski crowed was the “first intelligent design think-tank at a research university.” The conference itself, wrote Barbara Forrest and Paul R. Gross in their Creationism’s Trojan Horse (2004), “marked the Wedge’s official entry into the world of mainstream academia”: the idea was to invite world-class scientists and scholars to the conference so that the “intelligent design” crowd could bask in their reflected splendor. Dembski’s conference paper, “Can Evolutionary Algorithms Generate Specified Complexity?”—later published in Niels Henrik Gregersen’s collection From Complexity to Life: On the Emergence of Life and Meaning (2002)—contained the same description of the algorithm, almost verbatim.
Wesley Elsberry, then a graduate student at Texas A&M University and later a colleague of mine at NCSE, and a persistent critic of creationism throughout, wrote to Dembski in October 2000. He quoted Dembski’s description of the algorithm from the conference paper and commented, “Steps (2) and (3) appear to be inventions rather than descriptions. What is the basis for claiming that steps (2) and (3) represent Dawkins’[s] ‘weasel’ algorithm?” He received no reply, which is perhaps not surprising, since Dembski was then embroiled in a controversy at Baylor. Established over the protests of the university’s faculty in 1999, the Michael Polanyi Center remained controversial. After a review committee produced a report tepidly recommending its continuation, Dembski issued a gloating announcement of victory, angering the faculty and administration to such a degree that he was relieved of his duties as the center’s director.
Perhaps it was too much to expect Dembski, while he was busy shooting himself in the foot at Baylor, to answer Elsberry or to address the issue in the revised version of his conference paper published in the Gregersen volume. But well after that debacle, in No Free Lunch (2006), Dembski describes the algorithm as latching, using exactly the same words: “whenever an alteration happens to match a corresponding letter in the target sequence, leave it and randomly alter only those remaining letters that still differ from the target sequence.” And three years later, in “Conservation of Information in Search: Measuring the Cost of Success” (2009), coauthored with Robert J. Marks II, Dembski cited The Blind Watchmaker as using a procedure in which “characters are ratcheted into place.” This citation, moreover, occurred even though, after reading a prepublication version of the 2009 paper, Elsberry notified Marks about the error.
In 2009, on what was then his blog, Dembski tried to argue (on the basis of questionable evidence) that Dawkins’s algorithm was latching. But now, however, in Being as Communion, apparently abandoning the 2009 argument, he gets it right. Indeed, he reproduces a sample run (taken from a Wikimedia page) in which there is no latching: for example, a correct A in the twenty-fifth position in the seventh generation becomes an incorrect I in the eighth generation. There is no acknowledgment of his previous errors, though, and Elsberry is not mentioned in the book. To be sure, not much of substance depends on whether the algorithm of Dawkins’s example is interpreted as latching or not, although it’s interesting to compare implementations of latching and non-latching algorithms, as Ian Musgrave showed in a post at The Panda’s Thumb in 2009. But the story is perhaps instructive as to the quality of “intelligent design” scholarship.