If you've ever wrestled with the question of God's existence, it's time to knock it off and take up needlepoint, because someone has figured it all out. In his book The Probability of God, physicist Stephen Unwin has calculated the probability of the existence of a personal god to be 67 percent.

The Probability of God has its good points. Unwin explains his method -- a controversial technique called Bayesian probability analysis -- in clear, accessible terms, and includes an informal derivation of Bayes' theorem. He writes about his complex subject in an entertaining (if excessively jokey) and non-dogmatic style. He resists the temptation of intelligent design, refuses to lean on the argument from design, and recognizes the anthropic principle, although he displays an odd affinity for Pascal's Wager. I can't exactly recommend Unwin's book, but he does present a refreshing angle on the problem of God's existence.

I believe his method to be flawed, however. To begin, Unwin sets the baseline probability for God's existence at 50 percent, calling this an "expression of complete ignorance" and therefore an ideal, unbiased starting point for analysis. As a skeptic, I find this absurd. If I proposed an Bayesian analysis of the existence of unicorns or dragons, would Unwin agree with a baseline probability of 50 percent? I doubt it. In my own analysis, I used a baseline probability of 10 percent, which I believe to be mighty generous.

Unwin then identifies six "evidentiary areas," which are:

1. The recognition of goodness,
   
2. the existence of moral evil,
   
3. the existence of natural evil,
   
4. the existence of intra-natural miracles,
   
5. the existence of extra-natural miracles, and
   
6. the existence of religious experiences.

I have two problems here: why only six evidentiary areas? Surely we could think of dozens more, which would make for a more accurate analysis. And why these six areas? Four of the six areas appear to favor God's existence. To his credit, Unwin recognizes this weakness, admitting that "you may even have new evidentiary areas to add."

Unwin then assesses each area, and assigns a probability for that evidentiary area given the existence of God, and a probability for that evidentiary area given the nonexistence of God. For example, he decides that if God exists, the probability that man would recognize goodness is 100 percent. But if God doesn't exist, he guesses, the probability that man would recognize goodness is 10 percent. After assessing each evidentiary area, he plugs these figures into the Bayesian equation to generate an updated probability. After assessing all six evidentiary areas, he arrives at a "final truth probability" for God's existence of 67 percent. Well, maybe not "final." Unwin cheerfully acknowledges that "your assessment of the evidence may differ," and offers instructions for creating a spreadsheet to make your own probability estimations and calculations.

I ran through the analysis myself (without the spreadsheet, thank you) with different probability estimations and added another evidentiary area:

There is no direct, sensory evidence for God's existence, even though he loves each one of us very much and wants us to believe in him.

I calculated the final truth probability for God's existence to be 0.01 percent. I am Joe's Complete Lack of Surprise, because... I'm an atheist. I bet a devout Southern Baptist would use still different numbers, add even more evidentiary areas, and generate a final truth probability of 99.9 percent. This is, I think, the real weakness of this line of inquiry. Bayesian probability analysis may be valid and useful in solving some types of real-world problems, but when dealing with a completely subjective problem like the existence of God, the final results of this method will only mirror the prejudices and preconceived notions of the man punching the calculator keys. Therefore, Unwin's method is useless because it reveals nothing.

+posted by Lawrence @ 6/28/2004 11:08:00 PM