Cornelius Hunter has posted an odd argument:
Is there evidence for evolution? Sure, there is plenty of evidence for evolution. But there are significant problems with evolution. There is plenty of evidence for evolution just as there is plenty of evidence for geocentrism. But the science does not bode well for either theory.
So the evidence for evolution follows this general pattern: Even at its best, it does not prove evolution to be a fact. And furthermore, the evidence reveals substantial problems with evolution.
So how can evolutionists proclaim evolution to be a fact with such fervor? There seems to be a glaring mismatch between the evidence and the truth claims of evolutionists. The answer is that evolutionists use contrastive reasoning. Evolution is not claimed to be a fact based on how well it fits the evidence, but rather on how poorly the alternative fits the evidence. Evolution is proved by the process of elimination.
In other words, Hunter is arguing directly against Dembski:.
In eliminating chance and inferring design, specified complexity is not party to an argument from ignorance. Rather, it is underwriting an eliminative induction. Eliminative inductions argue for the truth of a proposition by actively refuting its competitors (and not, as in arguments from ignorance, by noting that the proposition has yet to be refuted)
Now, in my view, both Hunter and Dembski are wrong, but Hunter is not wrong because of Dembski’s argument, and Dembski is not wrong because of Hunter’s argument, or at least, not as it stands.
Hunter objection to “contrastive reasoning” is based on two misunderstandings, as far as I can see. The first is of the very text he quotes to support his argument, namely Eliott Sober’s Evolution and Evidence:
This last result provides a reminder of how important the contrastive framework is for evaluating evidence. It seems to offend against common sense to say that E is stronger evidence for the common-ancestry hypothesis the lower the value is of [the probability of E given the common-ancestry hypothesis]. This seems tantamount to saying that the evidence better supports a hypothesis the more miraculous the evidence would be if the hypothesis were true. Have we entered a Lewis Carroll world in which down is up? No, the point is that, in the models we have examined, the ratio [the probability of E given the common-ancestry hypothesis divided by the probability of E given the separate-ancestry hypothesis] goes up as [the probability of E given the common-ancestry hypothesis] goes down. … When the likelihoods of the two hypotheses are linked in this way, it is a point in favor of the common-ancestry hypothesis that it says that the evidence is very improbable.
In a related thread, Hunter wrote, in response to a comment of mine:
So you can see that according to the likelihood ratio, the argument for CA strengthens as p becomes smaller. And as p becomes smaller, the conditional probability for CA also becomes smaller.
In other words, the worse the probability for CA, the better the case for CA, because the SA probability got even worse yet.
The context here is that Sober is considering a scenario in which two species share the same genetic sequence, and comparing the hypothesis that the species share a common ancestor (“CA”) with the hypothesis that they have a separate ancestry (“SA”). He shows how the relative probability of these two hypotheses depends on the probability of the sequence arising spontaneously (“p”).
As I try to explain in my response to his comment, Hunter seems to have confused the probability for the data, given the hypothesis, with the probability of the hypothesis, given the data. The former indeed goes down as p goes down. However, as p goes down, the latter goes up. This seems to be a simple error on Hunter’s part, but it is an important one as it goes to the heart of his argument: the reason that the posterior probability of CA given the data goes up relative to the prior for CA, and therefore also relative to the posterior for SA, is that p(SA)=1-p(CA). If ancestry is not common then it is separate. There is no excluded middle.
Hunter has failed to note that in the case of a pair of hypotheses in which one is the null of the other, as the posterior probability of one goes up, the posterior probability of the other goes down; in this case, therefore, “contrastive reasoning” is completely valid, and Sober is right that if we find a sequence in two species with a very low p, that is extremely strong evidence that the hypothesis that they common ancestry is correct – the posterior probability of CA will approach 1, and the posterior probability of SA will approach 0.
He then asserts that Sober’s logic is applied (either by Sober or by some evolutionist unspecified) to a piece of reasoning in which there is an excluded middle:
Science cannot know all the alternative explanations for the origin of the species. When evolutionists conclude evolution is a fact via the process of elimination, they are making a subtle but crucial non scientific assumption—that they know all the alternative explanations.
But here, we are not comparing “Common Ancestry” versus “Separate Ancestry”, where p(SA) must = 1-p(CA), and we know the probability of the data, given SA, but Evolution against Not Evolution, where we simply cannot estimate the probability of the data given Not Evolution, because we cannot know what those those Not Evolution hypotheses are. Of course, even in Sober’s example, if the “SA” hypothesis includes a maverick God who inserts odd sequences into his critters for no good reason (Hunter postulates a deleterious one), then our calculations of the probability of the data given SA will be off, but this is not the point Hunter makes.
So IMO, Hunter has misunderstood Sober, and has also misunderstood evolutionist claims, but is nonetheless right to say that you cannot infer that a hypothesis is correct, merely because it explains the data better than any other hypothesis on the table. There may be a better one round the corner right now, and indeed, AFAIK, evolutionary scientists regard all evolutionary hypotheses as provisional, because that is pretty well the core of the scientific methodology – that all hypotheses are potentially falsfiable by a better one.
And that’s what makes Dembski wrong. All the arguments of Dembski’s I have read for inferring design assume that he has specified his null in such a way that the only alternative is Design. For Dembski, 1-p(Not-Design) = p(Design). Which would be fine if he actually calculated p(Not-Design|data). But he doesn’t. He calculates the probability of a very specific Non-Design hypothesis, given the data, which is not the evolutionary hypothesis, nor is it Design. In other words, he excludes an enormous middle, and Hunter would be absolutely right to reject Dembski’s inference as fallacious, by exactly the same reasoning as he rightly rejects the straw man claim that evolution is true because it has a greater probability of being correct than any possible alternative.
That is certainly fallacious, I agree. I look forward to seeing Hunter’s next post on how Dembski’s arguments are similarly fallacious 😉