Evolutionists on the web have questioned why my former doctoral supervisor at the University of Chicago, Leo Kadanoff (postdoc of Bohr, NAS member, winner of the Wolf Prize, Isaac Newton Medal, etc.), decided to invite me to speak about my work on information.

The answer can be found in an email from him to me prior to issuing the formal invitation to speak at the University of Chicago. I had sent Kadanoff a copy of my latest book, Being as Communion: A Metaphysics of Information, in draft form. After reading the first part, he remarked:

As you well know, information is the primary topic in 21st Century science. Even my own work on phase transitions is often phrased as how one part of a material communicates with another. Like matter, information flow can in some situations be guided by intelligence and in others can flow without the aid of a guiding thought.

Coupled with the sense that I might have something interesting to say about information to UofC grad students, Kadanoff therefore, in consultation with several other faculty, issued the invitation.

That might have been the end of it, except that a video of my lecture with slides interspersed was posted on YouTube. This led the University of Washington's Joe Felsenstein (pictured at right) to issue a rebuttal at Panda's Thumb, which has in turn been picked up by PZ Myers and Jerry Coyne as the official debunking of my talk.

Felsenstein's criticism of my talk and work constitutes a straw man, attributing to me an argument that he prefers I had made and that he can thus refute, rather than the one I actually make. I'll run through his most salient missteps:

It turns out that Dembski's current argument is based on two of his previous papers with Robert Marks.

Actually, in my talk, I work off of three papers, the last of which Felsenstein fails to cite and which is the most general, avoiding the assumption of uniform probability to which Felsenstein objects. That paper is William A. Dembski, Winston Ewert, and Robert J. Marks II, "A General Theory of Information Cost Incurred by Successful Search," in Marks et al., eds., Biological Information: New Perspectives (Singapore: World Scientific, 2013). This paper was clearly referenced in my slides.

They [i.e., these papers] involve considering a simple model of evolution in which we have all possible genotypes, each of which has a fitness.

False. Especially in light of the third paper, the way I represent search is perfectly general and can handle any mathematical model of evolution out there.

The fitness surfaces implicit in Dembski and Marks's argument are known as 'white noise' fitness surfaces.

This claim captures Felsenstein's main line of criticism, which is that we don't give smooth fitness surfaces sufficient play, making them compete with white noise surfaces, which in turn totally drown them out. Thus we fail to give smooth surfaces their proper due, and our argument is supposed to collapse. But in fact, our approach to search assumes nothing about "white noise surfaces" or even the way fitness is represented at all. As I noted in my talk, search at its most general involves "query feedback," which in evolutionary search takes the form of fitness. But the range of fitness functions to which my work applies is perfectly general. This should be evident from one of the examples I present, namely Dawkins's METHINKS IT IS LIKE A WEASEL simulation. All the fitness functions considered (by him and me) are unimodal and gradually increasing as proximity to the target increases. These are all as smooth as can be.

Dembski and Marks's argument is not actually an Intelligent Design argument. It argues that a Designer is needed to explain the shape of the fitness surface, but once that surface is smooth enough, natural selection and other evolutionary forces do the rest. So there is no Design Intervention needed.

Felsenstein is perhaps a quarter right here. Marks and I do think that insofar as evolutionary processes produce specified complexity, this is ultimately due to a designer fine-tuning the evolutionary process. But our actual work on Conservation of Information only shows that any evolutionary theory is necessarily incomplete and cannot account for the creation of the information that the evolutionary processes limned by the theory supposedly outputs. Moreover, we never assert that once the fitness surface is smooth enough, evolutionary forces do the rest. Smoothness by itself is irrelevant. One can have all sorts of smooth fitness functions that drive no interesting evolution. Indeed, in the very papers of ours cited by Felsenstein, "smoothness" comes up only in passing, and always as an example of a type of fitness function relevant in certain searches. Again, our approach is much more general than Felsenstein lets on.

Dembski portrays Dawkins as arguing that the Weasel model shows that natural selection can originate information ... Dawkins's model was a teaching example to show why creationist debaters who argue that natural selection is doing a "random" search are disingenuous.

Actually, what I show is that all evolutionary computing models fail to originate information, and I give Dawkins's model as an example, arguing that more complicated evolutionary computing models are at base no different from Dawkins's model. They all fail to originate information.

Because Felsenstein's critique bears no resemblance to what I was actually doing in my University of Chicago talk, let me summarize what I did say there.

Briefly, I started by assuming that if biological evolution is to be an exact science, then it must be possible to model it on search. I then considered search at its most general, laying out its key components. I then presented the key result of Conservation of Information, namely, that for any search space with a target of small probability p, if one wants a search that will find that same target with a probability q (greater than p), the probabilistic cost of finding such a search is at least p/q. What this means is that, at the end of the day, one hasn't gained anything because if finding such a search has probability p/q or less, and then once one has found such a search, one only has q probability of finding the target, then the total probability of finding the target with such a staggered search is still p or less. This result, I argued, holds with perfect generality. It does not assume anything about the nature of the fitness surfaces, or working off a full set of genotypes, or any other such limitation on search as Felsenstein suggests.

Let me urge fair-minded people who have read Felsenstein's criticisms to listen to my actual talk and then read the three papers cited. Alternatively, if such fair-minded individuals lack the technical background to appreciate these papers, let them read the last few chapters of Being as Communion, which summarizes the significance of these papers for evolutionary theory in a more user-friendly way.

Images: Stone figure, University of Chicago; Quinn Dombrowski/Flickr. Joe Felsenstein/University of Washington.