by Joe Felsenstein
Over at Uncommon Descent an unusual discussion has erupted. A commenter named “MathGrrl” who has been occasionally active there as a critic of ID has actually been allowed to make a guest posting. She gave several examples of situations where one could make a specification of what were the best genotypes, and asked how in these cases Complex Specified Information could be defined. She has handled the discussion with great restraint. Several hundred comments later no consensus has emerged. Commenters at anti-ID blogs ( here, here, here, here, and here), have concluded from this that the concept of CSI is vacuous.
I’d like to give a perspective that may be unpopular here. I don’t think Complex Specified Information is a vacuous concept, though we usually do not have enough information to actually calculate numbers for it.
Simply put, birds fly and fish swim. They do so a lot better than organisms coded by random strings of DNA (formed by mutation without natural selection – organisms coded for by monkeys typing with four-key typewriters). If we could imagine looking at all possible such organisms with the same length genome as (say) a bird, the fraction of them that would fly as well as a bird, or better, would be incredibly tiny. So tiny that if every particle in the universe were a monkey with an ATGC typewriter, there would not have been enough time by now to produce anything as good as a bird even once since the time of the Big Bang. That is the essence of William Dembski’s argument. Note that getting technical about information theory is not required. People love to contradict each other about information theory, but we can set most of that part of the argument aside.
A simple definition of Specified Information would be that it is the negative log (to the base 2) of the fraction of those sequences that are better at flying than a bird. We don’t have enough information to actually calculate it, but we can be sure that it is big enough to pass Dembski’s threshold of 500 bits, and thus CSI is present.
So am I saying that CSI is present in examples of life? Yes, I am. So does that mean that it follows that design is present in those cases? No, it does not. As I have explained before (here), Dembski draws the conclusion that the presence of CSI proves design because he has a theorem, the Law of Conservation of Complex Specified Information (LCCSI), which supposedly proves that an amount of specified information large enough to constitute CSI cannot arise by natural processes, even once in the history of the universe. In fact, he is wrong, for two reasons:
* His theorem is not proven. Jeffrey Shallit and Wesley Elsberry pointed out (here) that Dembski violated one of the conditions of his own theorem when gave his proof that this large an amount of SI could not arise by deterministic processes.
* In any event, to use his theorem (even if it were proven) to rule out natural selection you have to use the same specification (say “flies as well as or better than this bird”) both before and after evolutionary processes act. And this Dembski does not do. His conservation theorem involves changing the specification in midstream. When you require that the specification say the same, you can immediately see that the amount if SI cannot be conserved. Natural processes such as natural selection can improve the flight of birds.
Advocates of ID endlessly repeat the mantra that the presence of CSI proves that design is present. They are relying on Dembski’s LCCSI, whether they know it or not. But natural selection can put Specified Information into genomes, and when it acts repeatedly, can easily exceed the threshold that Dembski uses to define CSI. The issue is not CSI, it is the conservation law, one that has not been proven in any form that is relevant to detecting design.