The Dawn Treader

Dembski dedicates an entire chapter to this discussion in his excellent book, .  This chapter, btw, is available for free (in PDF).  Click to read the whole chapter.  No charge.

When Dembski speaks on probability, one better pay close attention.  He speaks with the authority of a Phd in mathematics from the University of Chicago.  He knows of which he speaks.

Basically, Dembski eschews the use of Bayes.  His design inference approach takes a different tack.  Dembski prefers detecting design by eliminating the chance hypothesis.

Bayesian design arguments, by contrast, don't eliminate a hypothesis.  They detect design by calculating higher probabilities for the design hypothesis vis a vis the chance hypothesis.

Why does Dembski not like Bayes?

It is too subjective.  It depends on prior probabilities, which are highly subjective and open to attack, and it depends on conditional probabilities ... which are highly subjective and open to attack.

While the formula works marvelously, if you use bogus probability estimates, you will get unreliable answers.

He is right about that, of course.  Garbage in.  Garbage out.

With Bayes, you can spend most of your time answering challenges to your prior and conditional probabilities -- instead of discussing how a particular hypothesis is strengthened or weakened by evidence.

Here is why I am sticking with Bayes ... for now.

I think it models the way we naturally believe things.  Beliefs are not binary things.  We hold some beliefs stronger than others.  Bayes reflects this.  The frequentist approach (Dembski's preference) of eliminating a hypothesis is not really how most people arrive at beliefs ... in my opinion anyway.

Bayes is hot these days.  It is hard to argue with success.  Pragmatically speaking, Bayes "stock" is on the rise.  It has proved remarkably successful in searching for things -- from searching on the internet, to searching for a lost H-bomb, to searching for a lost submarine, to searching for real email versus spam ... to searching for truth about important things like did a resurrection really take place 2000 years ago.

Bayes does not require accurate probabilities to prove a point.  Plantinga, Craig and Moreland all use Bayes to demonstrate a point -- and none of them overreach and argue dogmatically for their conclusion.  Apparently Swinburne, an Oxford professor, does use Bayes to argue that there is a 97 percent probability that the resurrection occurred.  That seems a little extreme to me.

Finally, I think Bayes is a useful approach to combine with cosmological ID arguments.  While it can be used in the fine tuning argument, I think it works particularly well in the rare earth version of cosmological ID.  The conditions required to find a host planet, host sun, host solar system and host galaxy to support life are numerous and extreme.  We actually have some pretty good estimates on the likelihood of those conditions.  How you say?  Thanks to the SETI project, we are looking for life in the universe quite rigorously.  Because of SETI, we have learned how to try to search intelligently -- and just how many conditions are required to host life.  With stronger probabilities come stronger estimates -- and in this case, a stronger hypothesis that the universe was designed rather than not designed.