RE: [asa] The Parsimony Principle (aka Occam's Razor) and atheism

I totally agree with your post. In particular, that "God is not part of the model." However, man is not totally part of the model either since one cannot detect his consciousness with purely physical devices. That is to say, there will never be a theory of man that does not include the supernatural. Science is done by collecting data with purely physical devices. However, man, as a "detector," can detect the supernatural, which can never be explained in purely physical terms. In fact, even life cannot be explained in terms of purely physical terms. Therefore, there will never be a complete theory of life.

Moorad

________________________________

From: asa-owner@lists.calvin.edu on behalf of Iain Strachan
Sent: Tue 7/10/2007 5:17 PM
To: AmericanScientificAffiliation
Subject: [asa] The Parsimony Principle (aka Occam's Razor) and atheism

I've noted that an appeal to "the most parsimonious explanation", or Occam's Razor is a frequently used argument by atheists to justify not believing in God.

The argument goes something like this. Because the laws of physics and the theory of evolution can explain, in principle, everything there is, including all the complexity of life, etc, there is no need to invoke a Creator to explain it all. A universe without a Creator is a simpler, and therefore more parsimonious, explanation than a universe with a Creator. Hence the god-less universe is to be preferred.

However, I think this argument is flawed, for the following reason. Occam's Razor, or the principle of parsimony is certainly to be used and encouraged in trying to decide between different scientific models. For example, I have a bunch of data and I have to fit a polynomial curve through it. What order polynomial should I choose? ( e.g. linear, quadratic, cubic, quartic etc). In general the higher the order the better the fit, but the more complex the model (requiring more coefficients). One good way of deciding this is to look at the data-fitting errors. One can describe the dataset as a set of polynomial coefficients plus the residual errors. The better the fit, the lower the residuals. For each model, one has a "data description length", consisting of the length of the model coefficients and the length of the residual errors. Smaller errors require less bits to transmit, so have a shorter description length, but at the cost of a more complex model. The optimal!
  trade off is to get the shortest description length, and this will be the most "parsimonious" description of the data. Such methods have a sound basis in probability theory, and can be shown to be equivalent to Bayesian inference, which is one of the key ways probabilistic models are derived nowadays. The simplest way to put this is that given a number of models each of which fits the data equally well, the most "parsimonious" model is the most probable. If anyone's interested, I can probably point to technical papers that describe this.

But it is this connection with probability theory that, to my mind renders the "parsimony" argument against God as invalid. Firstly, it can only really be valid if God is "part of the model" - the explanation of why things are as they are. This is the fallacy of the Intelligent Design position - the claim that evolution can't explain everything, and therefore one needs a Designer (aka God) to explain the complexity of life. That argument perhaps falls foul of Dawkins's arguments in "The God Delusion" - that the Designer is even more complex than we are and hence even less probable. I think that by requiring God to "explain" the existence of various complex artefacts, one is falling into the Occam's Razor trap.

But it seems to me that the obvious rejoinder is that God isn't "part of the model", but is traditionally viewed as transcendent over the model. Furthermore, one simply can't make probabilistic estimates about the existence of God. (Dawkins wants to, but can only do so by requiring God to be part of the model - how could such a complex entity come into being?). But the supernatural, by its very nature, can't be subjected to scientific measurement and laws (otherwise it would be natural and not supernatural).

About 20 years ago, in my church, I was asked by an old lady with crippled arthritic hands to pray for her hands to be healed. I was horrified at the prospect of effectively being asked to perform a miracle, but the lady was in quite a lot of pain and distress and there seemed no way out but to comply (though I wanted to run away like a wuss). So I went through the motions of what you're supposed to do - laid my hands on her hands and prayed for healing - not with much hope, I'm ashamed to admit! Well, her hands, which were crippled and stiff with pain, immediately freed up, to everyone's amazement (especially mine!). As I was driving off in the car park, I saw her - she lifted her hand and waggled her fingers at me to show that they could move where previously they could not.

But the point to make about this experience, which was real enough to me and to the lady herself, is that it could never be repeated in a laboratory under double-blind conditions, and so one couldn't possibly make any sort of probabilistic inference about such an event. No similar event has happened to me since then though I've heard of a few other anecdotal accounts of healings, just as the above is. But in science, and especially medicine, anecdotal accounts simply don't count as evidence - the effects have to be repeatable. I guess a more "parsimonious" explanation would be that she didn't have arthritis at all and that her paralysis was psychosomatic. But there is absolutely no way of checking out the probability of either explanation - and in order to invoke the principle of parsimony to distinguish between the two one MUST be able to estimate probabilities.

So in the end it would come down to prior beliefs. If one's prior belief is that there is no God, then one MUST opt for the psychosomatic explanation (or something else like saying she was faking it for some bizarre reason). But to someone who believes in a God who can perform miracles of healing, then it's not an unreasonable explanation to think that this was what happened. But to make an estimate of the probabilities of either event is, I believe, impossible & hence the principle of parsimony just doesn't apply.

Iain

To unsubscribe, send a message to majordomo@calvin.edu with
"unsubscribe asa" (no quotes) as the body of the message.
Received on Wed Jul 11 08:24:45 2007