This month's Skeptical Inquirer is devoted to "Science and religion"
(although most issues tackle these!). Interestingly, no arguments are
provided from a Bayesian perspective. If Bayesian reasoning is the
scientific method of dealing with degrees of belief, it's curious that the
scientific community is slow in adopting it.
There are clearly many conceptions of God, and therefore many hypotheses. A
number of contributors talk about the "Judeo-Christian-Islamic" God, which
is truly poorly specified---More or less monotheistic, more or less
demanding of certain ritual, more or less demanding of certain
belief---this conception of God is more Wittgensteinian family resemblance
than true definition. No wonder it doesn't pass the clarity test. (And why
are we excluding Eastern notions of deity?)
For this reason, I don't understand why Pascal's wager (okay, or Lasky's)
must have only two hypotheses. Such limitation is clearly a bad model. But
it's impossible to do with all the hypotheses that are out there. (There's
a nifty book called The Denominations of the United States---over 230
official religions in the US alone.)
I also don't understand the problem with "deciding to believe." Didn't the
UAI community deal with this problem over 10 years ago in going beyond the
logical AI folks by explicitly separating belief from action?
Doesn't any of our community's work have anything to add to these
philosophical musings of folks from other disciplines?
