If ever there was a seminar where lunch should have been provided, this was it

I've just listened to a talk over in the Biology department about the theories of William Dembski, the idea of "specified complexity" and the "no free lunch" theory. Dembski is the leading light of intelligent design. Following Dembski's arguments reminded me of something Joe Wright said about Antonin Scalia: you know that the conclusion is wrong but it's really hard to see what was wrong with the writing that got you there.

I left the talk with no idea what "specified complexity" is; only that it seemed to have many proposed definitions. When one of the definitions is discredited, intelligent design "theorists" can fall back on one of the other definitions. For me, this means simply that the concept is not falsifiable and thus obviously not science. It's obviously that neither intelligent design theorists nor biologists under information theory (I'm an electrical engineer and I'm not that great with it myself.)

The case of no free lunch is more difficult. The No Free Lunch theorem is a legitimate result (Wolpert and Macready, IEEE Trans. of Evolutionary Computation 1 (1), April 1997) The paper discusses optimization algorithms. Given a space X consisting of a finite number of points, assume that is a randomly distributed cost value Y for each point. The No Free Lunch Theorem says (very very informally) that all optimization algorithms are equally good on such a space. What this implies is that if a particular algorithm is better than just picking points at random for some subclass of problems, there must be another subset of problems where it is worse than picking points at random.

Dembski interprets this theorem to mean that no "evolutionary search" algorithm can be better than random selection. And since the probability of a random mutation being beneficial to an organism is practically zero, no evolutionary search algorithm can work because the probabilities of success is too well. Ergo, evolution is impossible; ergo, an intelligent designer is the only explanation.

This argument misinterprets Wolpert and Macready's result. That result only says that an evolutionary search won't work for some problems. It doesn't say that it can't work well for the problem of finding mutations that improve organism (aka evolution) unless you assume that the optimal mutations are randomly scattered, without any correlation, over the set of all possible gene sequences. There's not a lot of understanding at this point as to how much correlation or how little correlation there is between beneficial gene sequences.

So Dembski's model is based on as assumption about the lack of correlation of beneficial gene sequences. This is how a scientist works with a model:

1. Make some observations.
2. Build a model (mathematical or otherwise logically consistent) in an attempt to explain these observations.
3. Use the model to make predictions as to what other phenomena should occur in nature.
4. Make more observations.
5. If the phenomena predicted by the model are observed, your model is good. If not, you need to refine or replace your model.

Dembski doesn't work like that. He built a model and charitably it's based on observations of natural processes. He makes predictions based on this model and based on those predictions, not only should natural selection be impossible, but also observed phenomena like virus and bacteria gaining resistence to drugs.

Now from this you can infer that your model is wrong or that nature is wrong. Obviously there's something wrong with his model since it implies the impossibility of observed phenomena. The correct thing to do, assuming the No Free Lunch model was presented in good faith, is to refine the model to explain how these phenomena could have occurred.

But the problem is that the No Free Lunch model gives an answer that a lot of people want to hear because it doesn't interfere with their religious beliefs. At this point intelligent design is not science because the philosophy of the intelligent design movement is to maintain that the conclusion is correct regardless of the observations. This is exactly what they accuse the "Darwinists" of doing, insisting their theory is correct while ignoring evidence to the contrary. This projection seems typical among much of the right wing in the US.

Is there a controversy in scientific terms here? No. These theories start out following the form of science but stubbornly hold on to the conclusion they want in the face of evidence to the contrary. Scientists are sometimes guilty of that, but the progress of science pushes these faulty conclusions aside.