Last March Tom English and I posted an argument here here at Panda’s Thumb analyzing an argument by William Dembski, Winston Ewert, and Robert Marks. They had made an argument that evolutionary “search” would not do better than blind search; we proved that their argument showed no such thing.
In response to our analysis here of the Dembski-Ewert-Marks paper, Winston Ewert has replied at Evolution News and Views. As that site does not allow comments, I have finally gotten around to posting a response here (six months late). Tom has now put up a related thread at The Skeptical Zone; I will try to comment in both discussions.
Ewert rather dramatically reveals that Tom and I do not actually disagree with any of the theorems in their paper. And he’s right about that. How did they discover this remarkable fact? Perhaps it was by reading our post, where we said
We’re not going to argue with the details of their mathematics, but instead concentrate on what in evolutionary biology corresponds to such a choice of a search.
or by reading a comment in that thread where I also said:
As theorems they may be mathematically true, but the average poor performance of searches is true only because so many irrelevant and downright crazy searches are included among the set of possible searches.
Ewert is right that we did not question their theorems. Instead we concentrated on what would follow from their theorems. We showed in a simple model that once there are organisms that reproduce, with genotypes that have phenotypes and fitnesses, that evolution will find higher fitnesses much more effectively than random guessing. So is it true that having what they call Active Information, embodied in a fitness surface and in a reproducing organism whose genotypes have those fitnesses, requires that there be Design Intervention to set up that system?
The issue is not the correctness of their theorems but, given that they are correct, what flows from them. Dembski, Ewert, and Marks (DEM) may object that they did not say anything about that in their paper.
We don’t think that it is a stretch to say that DEM want their audience to conclude that Design is needed.
Let’s look at what conclusions Dembski, Ewert, and Marks draw from their theorems. There is little or no discussion of this in their paper. Are they trying to persuade us that a Designer has “frontloaded” the Universe with instructions to make our present forms of life? Let’s look at what Dembski and Marks have said about that (below the fold) …
Our analysis of what kinds of “searches” would be achieved by a reproducing organism that has fitnesses led us to this simple conclusion (summarized in a comment of mine in the thread following our post):
Their space of “searches” includes all sorts of crazy searches that do not prefer to go to genotypes of higher fitness – most of them may prefer genotypes of lower fitness or just ignore fitness when searching. Once you require that there be genotypes that have different fitnesses, so that fitness affects survival and reproduction, you have narrowed down their “searches” to ones that have a much higher probability of finding genotypes that have higher fitness.
In addition, the laws of physics will mandate that small changes in genotype will usually not cause huge changes in fitness. This is true because the weakness of action at a distance means that many genes will not interact strongly with each other. So the fitness surface is smoother than a random assignment of fitnesses to genotypes. That makes it much more possible to find genotypes that have higher fitness.
In short, with their theorems, Design is not needed to explain why a reproducing organism whose genotypes have fitnesses might be able to improve its fitnesses substantially. Just having reproducing organisms, and having the laws of physics, gets an evolving system much farther than a random one of DEM’s “searches”.
But here’s what William Dembski said about this in his interview at the Best Schools website:
The term “evolutionary informatics” was chosen deliberately and was meant to signify that evolution, conceived as a search, requires information to be successful, in other words, to locate a target. This need for information can be demonstrated mathematically in the modeling of evolutionary processes. So, the question then becomes: Where does the information that enables evolutionary searches to be successful come from in the first place? We show that Darwinian processes at best shuffle around existing information, but can’t create it from scratch.
I see this work as providing the theoretically most powerful ID challenge against Darwinian evolution to date. As for the attention this work has garnered, there has been some, but Darwinists are largely ignoring it. I’m justified in thinking this is because our methods leave them no loopholes. We’re not saying that evolution doesn’t happen. We’re saying that even if it happens, it requires an information source beyond the reach of conventional evolutionary mechanisms.
and here’s what Robert Marks said at his Evolutionary Informatics Lab website (http://evoinfo.org) (copied 7 Nov 2015):
By looking to information theory, a well-established branch of the engineering and mathematical sciences, evolutionary informatics shows that patterns we ordinarily ascribe to intelligence, when arising from an evolutionary process, must be referred to sources of information external to that process. Such sources of information may then themselves be the result of other, deeper evolutionary processes. But what enables these evolutionary processes in turn to produce such sources of information? Evolutionary informatics demonstrates a regress of information sources. At no place along the way need there be a violation of ordinary physical causality. And yet, the regress implies a fundamental incompleteness in physical causality’s ability to produce the required information. Evolutionary informatics, while falling squarely within the information sciences, thus points to the need for an ultimate information source qua intelligent designer.
In case there is any doubt, here is what Marks said in an “ID The Future” podcast (This is available at Marks’s Evoinfo.org website, but for some reason not at the Discovery Institute’s “ID The Future” website).
[Minute 13:28] Luskin: What is Active Information, and why does it point to the need for Intelligent Design to solve a problem, rather than an unguided evolutionary process?
Marks: Well the idea actually goes back to Bill Dembski’s book entitled “No Free Lunch”, which shows that, remarkably, if one is a doing a search and designing something, then one search, or one process, is on average as good as any other process if you have no idea about the problem you’re solving, in other words, the search is undirected – and that its blind search.
[I omit his Wierd Al Yankovich blind Rubik cube scene description. JF]
Marks: That’s an example of a blind search. And without information to guide you where you want to go, one search is as good as another search, which on the average is as good as blind search.
[I omit banter with Luskin about liking stupid movies and stuff about algorithmic information theory pioneer Gregory Chaitin liking their paper. JF]
Luskin: … Well, we appreciate the work that you’re doing and the papers that you’re publishing analyzing many of these evolutionary algorithms and asking whether they support a Darwinian view of life or an Intelligent Design view of life.
“Dr. Robert Marks, Active Information in Metabiology” 2014-05-31 16:00. Discovery Institute, Center for Science and Culture, Copyright Discovery Institute, 2014
Ewert himself, in his reply, uses the evolution of birds as an example:
While some processes are biased towards birds, many others are biased towards other configurations of matter. In fact, a configuration biased towards producing birds is at least as improbable as birds themselves, possibly more so.
Having postulated Darwinian evolution, the improbability of birds hasn’t gone away; we’ve merely switched focus to the improbability of the process that produced birds. Instead of having to explain the configuration of a bird, we have to explain the configuration of a bird-making process.
This example leaves it unclear what the “process” is. The reader may be tempted to conclude that it is the process that models an evolving population. And then the reader may think that if this evolutionary process succeeds in improving fitness, that some outside force is needed to set up the process so that it succeeds. But for their theorem to apply, the processes considered must include processes that make no sense as models of evolution. Processes that wander around among genotypes randomly, without being more likely to come up with higher fitnesses. Even processes that prefer to find genotypes with lower fitnesses. All of those are among the processes that must be eliminated before we get to processes in which genotypes have fitnesses, and those fitnesses affect the outcome of evolution.
But after narrowing the searches down to those that have organisms with fitnesses, the probability of success is much improved. And no designer needs to intervene to get that improvement.
All that occurs even when the fitness surfaces are extremely rough, with nearby genotypes having wildly different fitness. But the laws of physics strongly suggest that fitness surfaces will be much smoother than that. That smoothness is a further reason for evolutionary processes to succeed far better than blind searches.
In his reply, Ewert invokes the smoothness of the fitness landscape, and considers the smoothness to result from “laws or self-organization”
It is not sufficient to invoke the three-fold incantation of selection, replication, and mutation. You must also assume a suitable fitness landscape. You have to appeal to something beyond Darwinism, such as laws or self-organization, to account for a useful fitness landscape.
He does not seem to realize that those “laws” might simply be the laws of physics, and that the “self-organization” can simply be self-reproduction, something that all organisms do.
By simply declaring that we do not disagree with their mathematics, Ewert is avoiding the issue of what flows from those theorems. In their subsequent statements, Dembski, Ewert, and Marks have made clear that they regard those theorems as evidence for Design.
It is clear from these examples that Dembski and Ewert mean their theorems to be read as evidence for an Intelligent Designer either frontloading the evolutionary process, or for an Intelligent Designer intervening in it. But Tom English and I have shown that their Active Information can come about without that. It can come about simply by having a reproducing organism which has different genotypes, which have different phenotypes, and these have different fitnesses. And further Active Information can also come about by the predisposition of the laws of physics to bring about fitness surfaces smoother than “white noise” fitness surfaces.
Could that Active Information be enough to explain the evolution of, say, a bird? Do they have some argument that further “configuration of a bird-making process” is needed beyond that? There is actually nothing in their argument that requires that there be further Intelligent Design.
This is bad news for the implications that they have drawn from their theorems.
To say that is to disagree fundamentally with Dembski, Ewert, and Marks, however little we have questioned the validity of their theorems.
As usual, I will “pa-troll” this thread aggressively and send off-topic comments by our usual trolls (and replies to those comments) to the Bathroom Wall.