The Probability of God

In Richard Dawkins‘s new book “The God Delusion” (I’ve just finished reading it – review coming soon), he argues that Agnosticism is an irrational position, because it basically has to maintain that the probabilities for the existance and non-existance of some god are (at least roughly) equal. Dawkins argues that they are not. This has prompted many agnostics to come out and defend their positions.

One of them is John Wilkins, who argues quite well that one cannot set up probabilities in the first place:

The philosophical agnosticism I adhere to does not say anything about probabilities at all. It says, instead, that nothing can count for or against either position decisively. Probabilities are based in this case on prior assumptions – one uses Bayes’ theorem to determine whether or not the hypothesis under test is likely to be true, given other assumptions we already accept. And here is where the problem lies – which assumptions? To adopt and restrict one’s priors to scientific assumptions is question begging. You in effect eliminate any other conceptual presuppositions from being in the game. This has a name in philosophy – positivism. It is the (empirically unsupportable) claim that only scientific arguments can be applied. As Popper noted, this is self-refuting. You cannot prove the basic premise of your argument that only provable (or, let’s be generous, supportable) claims should be accepted. As this is not a supportable claim in itself, you have contradicted your own position.

I’m not sure that it’s impossible to set up probabilities of any kind. For example, consider the following hypotheses:

  • H_0: Our universe, U_0, has no creator
  • H_1: Our universe, U_0 has exactly one creator. She lives in another universe, U_1 and runs U_0 as a computer simulation. U_1 has no creator.
  • H_2: Our universe, U_0 has exactly one creator, living in U_1 and running U_0 as a computer simulation. U_1 has exactly one creator, living in U_2, running U_1 as a computer simulation. U_2 has no creator.
  • … and so on until:
  • H_inf: There is a (countable) infinity of universes U_0, U_1, U_2, … Our universe is U_0. For each universe U_i the following holds: U_i has exactly one creator, living in U_i+1 and running U_i as a computer simulation.

All of these are possible hypotheses about the creation of our universe, but since we have no empirical data nor any viable assumptions outside our own universe on any of them, Wilkes holds that we must be agnostic about all of them, unable to pronounce one more likely than the other. But can we really not say that H_32589 is less likely than H_1?

As I said, I’m not sure. Even if we could, the general question of whether there is any god (for my purposes defined as creator of our universe, whether omnipotent or not) is an infinitely more complicated one and I tend to agree with Wilkes that from our point of view it’s impermissible to reason about probabilities for or against it.

Given the last point, I should call myself an agnostic in Wilkes’s sense. Still, I much prefer to call myself an atheist, because I firmly reject the gods postulated by popular religions – gods not content with merely creating but gods performing miracles, listening to and answering prayers, and caring about us (often with the provision that we believe in them) and our (non-existing) souls. But also because it pisses some people off more ;-)