I'm reading Why Darwin Matters by Michael Shermer (it's an early Christmas present). Shermer is the editor of Skeptic magazine, and I'd heard about him, but never read any of his stuff. Unfortunately, I'm not too impressed with his writing.

Here's an early passage:

If the theory of evolution is so profound and proven, why doesn't everyone accept it as true? One source of resistance is the confusion over the verbs "accept" and "believe." I use the verb "accept" instead of the more common expression "believe in" because evolution is not a religious tenet, to which one swears allegiance or belief as a matter of faith. It is a factual reality of the empirical world. Just as one would not say "I believe in gravity," one should not proclaim "I believe in evolution." But getting hung up on the idea that one is supposed to "believe in" evolution just as you "believe in" God is just one brand of resistance to evolution.

This is partly an attempt to differentiate types of belief, and it doesn't make a whole lot of sense until you get further into the book, to a section entitled "Why Science Cannot Contradict Religion" and Shermer goes on to advocate what he calls "The Separate-Worlds Model" of science and religion, which is basically what Stephen Jay Gould was talking about when he referred to science and religion as "non-overlapping magisteria". It's basically a "separate-but-equal" view of science and religion. Here's Shermer's explanation:

Believers can have both religion and science as long as there is no attempt to make A non-A, to make reality unreal, to turn naturalism into supernaturalism. Thus, the most logically coherent argument for theists is that God is outside time and space; that is, God is beyond nature-super nature, or supernatural-and therefore cannot be explained by natural causes. God is beyond the dominion of science, and science is outside the realm of God.

So what's wrong with this view of science and religion? It raises the question of why you should believe in something that is "outside space and time". Pretty much any fantastical, fictional entity fits that description, so couldn't this argument be used to justify not only any religion, including polytheism, animism, etc., but belief in leprechauns, unicorns, and dragons?

Another problem is that most believers don't relegate their religion to the non-empirical world. They make empirical claims. Most Christians believe that Jesus was born of a virgin (which is an empirical question), that he was resurrected from the dead (another empirical question), and that he performed all sorts of miracles in-between. These are all statements about the real world. I guess as long as someone doesn't make these claims, then science and religion can stay in their nice little boxes and play nice, but Shermer is describing a form of deism, a non-personal god who doesn't interfere with worldly affairs, and I don't see many deists running around.

But back to the first passage, and belief. I really don't like the way he uses the words, though it becomes clear why he does later on. I'm perfectly fine using the word "believe" for any representation of the world in a cognitive system. I don't think it sounds particularly weird to say "I believe in gravity" or "I believe in evolution".

A couple of semesters ago I took a class called "The Computational Basis of Intelligence," in which we studied probabilistic approaches to robot perception and control. We used this book. The thing is, the probabilistic approach to robotics has become extremely successful, and the basic approach is to use statistical approaches to calculate and represent the robot's beliefs. Yes, they use the word "belief", and it not only fits conceptually, but if you're using Bayesian statistics, it's really the only appropriate word to use. The idea is that there's uncertainty and noise in every sensor system, whether it's vision or sonar or laser-guidance, and there's noise in all motor output (real gears grind together, there's friction, etc.). So for localization (i.e., for the robot to try to figure out where it is in the world after it has just moved) you represent its belief in its location as a probability distribution. Think of it as a cloud of points plotted on a 2D surface. The ones at the center are darker, representing more certainty in that position, and the ones near the edge are lighter, representing less certainty. The spread of the points is also a reflection of the degree of uncertainty. With a tight spread, you're fairly certain where you are. With a huge spread, there is a lot of uncertainty.

Anyway, it's perfectly acceptable to talk about cognitive representations, no matter what they are, as beliefs. And there are different ways of acquiring, testing, and revising beliefs. Some are much, much better than others.

So I think religious beliefs and scientific beliefs shouldn't be divided into two camps, like Shermer suggests. A belief is a belief. What does differentiate them is how they are acquired, tested, and revised.