Wednesday, February 25, 2009

Naturalism defeated or supernaturalism disconfirmed?

This is a question for those of you who have greater facility with Bayes' Theorem than I do. (In other words, this is a question for just about everyone.) Let's suppose that either theism is true or evolutionary naturalism is true and to keep things simple let's say that neither hypothesis is more likely than the other. (This is a gross simplification, but let it slide to keep the calculations simple. The second assumption is an assumption Plantinga makes in his initial presentations of the EAAN.) Let 'T' stand for theism, let '~T' stand for evolutionary naturalism, and let 'R' stand for the proposition that our faculties are reliable. Suppose for the sake of this discussion that Plantinga is right that:

(1) P(R/T) > P(R/~T).
(2) P(R/T) is high.
(3) P(T) = P(~T).

Given these three assumptions, doesn't evidence against R count as evidence against T? Playing with the numbers just a bit, suppose that we say that P(R/T) is .9 and P(R/~T) is .2. If we further suppose that P(T) = P(~T), P(~T/~R) is .8889. It seems that there is some evidence that we are unreliable in certain kinds of cognitive tasks that I'd think that we wouldn't expect ourselves to be particularly good at given the background assumption of evolutionary naturalism but would expect ourselves to be good at given the background assumption of theism. Having a bit too much fun with trolleys and chalk the other day, I discovered that the vast majority of my students start judging that they ought to engage in actions that they themselves deemed to be murderous only weeks earlier when we have 4 option trolley cases rather than the 2 option cases I initially presented to them. I started having my dose of Darwin's Doubt!

Now, I know there's the worry that if you accept ~T and P(R/~T) is low, you are supposed to acquire a defeater for everything you believe. But, it seems that this problem is easily avoided. I think Sober pointed this out in an article some years ago. We should focus on the reliability of certain kinds of methods and processes for discovering the truth instead of focusing on reliability across the board. There's ample evidence against reliability across the board but little evidence for the proposition that every faculty we have is unreliable. It would be self-defeating (perhaps) to believe that every faculty is unreliable, but not to believe that we are not reliable across the board. That we are unreliable moral judges does not mean that we oughtn't trust our scientific judgments. But, here's the kicker. This sort of move seems helpful to the naturalist but I don't think that the theist wants to say that the conditional probability that we are good moral judges on the assumption of theism is low or inscrutable.

Anyway, I think this is interesting. It seems that if the assumptions Plantinga often introduces in the course of setting up the EAAN he might have given the naturalist some assumptions to use to construct an empirical argument for God's non-existence. Roughly, we should expect to be competent moral judges if God exists but not if evolutionary naturalism is true. Given evidence of incompetence and the initial use of the principle of indifference to justify the starting assumption that P(T) = P(~T), we ought to decrease our confidence in theism accordingly.

I have no confidence in the above, by the way. I'm sure there are details that need to be ironed out, assumptions that need to be added in, and responses that are so damn obvious that any fool could see them. That's what the comments box is for.


dtlocke said...

I wonder what the evidence for our being incompetent moral judges is? I suspect the situation is this: we often make moral judgments that, upon sufficient reflection, we see to be incorrect. But if that's the case, we aren't really incompetent moral judges---we're just incompetent at making accurate moral judgments in a hurry. But surely the theist can respond that the conditional probability of that given T is not lower than the conditional probability of that given ~T.

btw, congrats on your new job!

Clayton said...

Hey Dustin,
Good question. Part of what led me to write this was the fact that it seemed surprising to discover that P had actually committed himself to a set of propositions that would enable us to use empirical evidence to disconfirm the hypothesis that God exists. (Discovered might be too strong.) You're right that I haven't produced that evidence but it seems we have some idea how to go about trying to gather that evidence. You are right that we have to be careful to distinguish the reliability of moral judgment from the reliability of moral judgment in a hurry.

Along with the trolley example I mentioned, I remember hearing about a study on Radiolab having to do with subjects asked whether they would smother a baby while hiding from troops who, if they were to discover them, would execute the lot of them. There was, apparently, something close to a 50/50 response to this sort of scenario and if we did the studies carefully it is quite possible that this sort of disagreement might not go away. And that would be troubling, I think. Done right, I wouldn't be surprised to discover that people are unreliable in their moral judgments. But, then again, I'm an atheist.

Thanks on the congrats, I'm really excited to be employed at all these days.


Mike Almeida said...


I don't think unreliability will cut in clear ways against the conditional probability of theism. For one small argument, note that (1) is true,

1. Pr(T/R) =

Pr(T) x Pr(R/T)/Pr(R).

Suppose the probability of theism is .5 for you (but it really doesn't matter much), and you agree with Plantinga that Pr(R/T) is pretty high: i.e., we'd be pretty reliable if God existed. Say it is .9, to give it some number. In that case Pr(T/R) = .45/Pr(R). Of course Pr(R) has to be at least .45. But as you increase the unconditional probability of our cognitive reliability (i.e., our reliability independent of God's existence) the conditional probability of theism on reliability decreases. So, I think you want to show that the unconditional probability of our reliability is high, not low. For instance, if Pr(R) goes to 1 (we're perfectly reliable), then R actually disconfirms theism.

Anonymous said...

‘I think Sober pointed this out in an article some years ago. We should focus on the reliability of certain kinds of methods and processes for discovering the truth instead of focusing on reliability across the board.’

This may be what theists try to do, claiming to have knowledge of divinity. It might be good for theists to be able to claim that their faculties are more reliable ‘across the board’ compared to the faculties of non-theists, but there seems to be no evidence that this is the case. If the only thing at stake is the reliability of the faculty involved in God-detection then the argument seems circular and the explanation ad hoc.

It’s not clear what ‘theism’ may stand for so as to be a theory that can compete with evolution. I’m not sure if a Gnostic who thinks there’s a God distinct from the creator would count as a ‘theist’ in this context.