To quickly explain, Dembski’s method is impractical because eliminating the alternative explanations is usually not feasible. It will work in trivial cases where the explanations are easily listed and the probabilities can be calculated without too much work. But really, is there any non-trivial case where that actually applies ?
That’s why pretty much nobody bothers to use Dembski’s method. To the best of my knowledge even Dembski hasn’t managed to apply it to biology. At one point, I recall, he was using the dodge that if the probabilities couldn’t be calculated the explanation should be discarded anyway. Hardly rigorous mathematics!
Yes, I think the usual method of detecting design is to consider a positive design hypothesis and use inference to the best explanation. Purely negative arguments are an unreliable method - it’s too easy to miss possibilities.
However Dembski’s method is supposed to deal with identifying the pattern in retrospect. That’s the point of specification - the observed pattern is shown to be “special”. The problem Dembski missed at the time is that there are plenty of “special” patterns and the probability of getting any one of them is (usually) what you really need. I’ve seen writings where he tries to correct for this by calculating the number of possible specifications but it’s a really bad omission.
The really weird thing though is that Dembski tried to sell his “Complex Specified Information” as an anti-evolution argument. But you can’t count any feature as CSI without ruling out evolutionary explanations. It makes me suspect that Dembski had a quite different argument in mind but couldn’t get it to work.
If you look at Mt Rushmore (a favourite ID example) you don’t go “I can’t think of a natural explanation, it must be designed” - but you might well go “humans carved that to honour famous people”
That is, Dembski’s method leaves design as an unexamined default. If you can’t think of a possible natural explanation for a pattern you are supposed to conclude design - even if you don’t have any idea at all of who might even want to do it.
In contrast most people would think of an actual explanation involving design and compare that to the evidence.
quote: I'm at a loss for how it is supposed to do this. Complex Specified Information makes sense to me, but he doesn't seem to propose a method to detect any specifications, they must be assumed.
The basic idea is pretty common. Given a fair coin the probability of any sequence of tosses depends only on the length. Obviously it would be daft to say that there must be something odd going on just because you’ve tossed a coins twenty times in a row. But if you get twenty heads in a row - which is a million-to-one shot you would be justified in thinking that maybe there is something making it come up heads.
Dembski does have some ideas on what makes a specification valid and they aren’t too bad, but I don’t think that they are objective enough to call his design inference a mathematical proof. I may quote them later - I do have a copy of the book (remaindered - no way I’d have paid full price)
quote: Evolutionary mechanisms are ruled out a priori with his method not because of the 'necessary' sieve in his Explanatory Filter, but because natural selection doesn't specify positively any patterns of surviving organisms genome, right?
Evolutionary explanations are supposed to be ruled out based on probability calculations. In reality I’ve never heard of Dembski coming close to doing that - in one case the “reason” was “Behe said so” which doesn’t even seem to be true. (In Darwin’s Black Box Behe admitted that there were “indirect” routes of evolution that sidestepped his argument. He assumed that they were too unlikely to count but never made an argument to that effect, and certainly never showed that the probability fell below Dembski’s probability bound. He was almost certainly wrong, too)
quote: but natural selection is the environmental pressure selecting negatively (akin to Dembski's sieve) the organisms which are unfit, and the rest of the population it is blind to, right? In other words, natural selection is simply death of the unfit, no?
That’s not really accurate. Positive selection for beneficial traits is an important part of evolution. The simplest explanation is that, in aggregate, genes that help individuals successfully reproduce become more common while genes that hinder that become less common.
quote: From my surface understanding of the Design Inference, it is nothing more than a Designer of the Gaps argument backed by bad math. His main thesis seems to be "it is too improbable, therefore magic"
Formally it is more “everything else I can think of is proven too unlikely therefore some unspecified designer did it by unspecified means”.
That’s poor as science (since the conclusion is too damn vague), risks error (it’s too easy to miss something) and pretty much useless where he wants to apply it. (The calculations for evolution would be a ridiculous amount of work even if we had good figures)
quote: Of course, internally his whole thesis hinges on getting the probabilities right, and from what I have seen he uses the Sharpshooter fallacy where probabilities are calculated after an event has occurred (which is the wrong way to do it).
In my experience it’s more a case of he finds excuses not to do the calculations or does the wrong calculation entirely. He does actually take some steps to avoid the sharpshooter fallacy, but having a more or less valid methodology - in principle - is no good if the application is always seriously lacking.
quote: In a nutshell, if we look at evolutionary pathways down to the fine grained level of individual mutations then every evolutionary pathway is extremely improbable
Yes, that is a potential flaw of the method but I haven’t seen Dembski do a good enough job of applying his method for that to be a real issue in practice. If Dembski ever gets to calculating the probabilities of evolutionary paths (and he may have, I haven’t looked at his work in the last few years) that might be an issue. But I haven’t seen him even try it, even when he should.