Wesley Salmon, Probability and Design Arguments

Wesley Salmon has argued that the design argument for God's existence is unsuccessful [1]. He writes:

We have considered the hypothesis that the universe is the product of intelligent design, and we have seen that a very rough estimate, at least, can be made of its probability by applying Bayes's theorem. The result is, of course, disappointing to the theist (or deist), for the conclusion is that the hypothesis is quite improbable.

Here's how he got there. Let E = x exhibits order and H = x is intelligently produced/caused. The relevant probabilities to consider are:

• Pr(E|H) (the probability that x exhibits order given that it is intelligently produced/caused)
• Pr(E|~H) (the probability that x exhibits order given that it is not intelligently produced/caused)
• Pr(H|E) (the probability that x is intelligently designed given that it is orderly)

Whether E supports H depends on whether Pr(E|H) is greater than Pr(E|~H). Is it? To answer this question, salmon has us consider the pool of things that exhibit orderliness. I like to think of this pool as an urn of balls, each ball representing an orderly thing that is either intelligently produced or not. Later, Salmon is going to have us draw randomly from this urn and see how likely it is to pull out something that is order and designed. Within the urn, we know that some order-exhibiting things will have been produced by intelligent agents. Houses, boats, engines, computers, watches and so on, are all instances of orderly and intelligently produced things. But Salmon invites us to consider the other members of the urn -- things which are not produced by intelligence but by "mechanical" or biological causes. There are a vast number of these things: atoms and molecules, galaxies, snowflakes, diamonds and other crystals, plants and other biological processes. All of these are orderly things/events which are not caused by "mechanical" or biological causes, rather than intelligence. A little reflection will show that the number of orderly things that are mechanically/biologically caused rather than intelligently caused is vastly greater than the number of orderly things that are intelligently caused. Salmon writes: 

When we consider the number of organisms, animal or vegetable, living on the earth -- including the millions of microbes which inhabit each human body -- we see immediately that biological generation operates in large numbers of instances...We now have evidence to indicate that [the universe] contains perhaps 10 billion galaxies, and that each of these contains perhaps 10 to 100 billion stars. And who can say how many atoms have been formed in the interiors of these stars? Our earth alone contains something like 10^50 atoms...[Hence,] There is...a great deal of evidence pointing to the conclusion that the number of instances in which mechanical causation operates is almost incomprehensibly greater than all of the rest [of the types of causation -- including intelligent/personal causation--] combined...If we count the relative frequency with which such entities were the result of intelligent design, we easily see that it is rather low. This conclusion holds even if we exclude such items as galaxies and atoms on the ground that we are not very sure how they are created. There is still a vast numerical preponderance of such occurrences as animal reproduction, growth from seeds, formation of crystals, and spinning of spider webs over the relatively few instances in which watches, houses, and ships are built by people.

So in the pool consisting of orderly things, there is a vastly greater proportion of orderly things that are non-intelligently produced to orderly things that are intelligently produced. If we draw randomly from the urn, it is exceedingly more likely that we'd reach in and pull out an orderly object that is not intelligently produced. Hence:

Pr(E|H) < Pr(E|~H).

And, hence, order is not evidence for intelligent design. Moreover, these rough estimates plugged into Bayes's theorem lead Salmon to conclude that Pr(H|E) is really low.

___________________________________________________________________________
[1] Salmon, "Religion and Science: A New Look at Hume's Dialogues" (1978).
[2] This conclusion also involves considering the prior probability of H relative to our knowledge of the sorts of things that are brought into existence. He argues that Pr(H) is really low.