Why do you expect happiness-causing beliefs to have a relationship to truthful beliefs?
The point is that if religion does make people happier, this makes it more probable than before on naturalism that lots of people would be theists, hence weakens the evidence for theism that comes from the surprising-ness of lots of people being theists. In other words, theism’s making people happier helps screen off the truth of theism from the phenomenon of widespread religious belief.
Of course, to be evidence against theism, the happiness thing has to lower the probability of theism on all of the evidence, not just lower it on one portion of the evidence. Some theists may want to argue that theists being happier is itself more likely on theism than on non-theism, and this isn’t terribly implausible. Without having a healthier sense of the conditional probabilities involved than I actually do, I don’t know how to evaluate the overall effect on theism’s posterior probability.
Actually, I should be more clear. Let T = theism, H = theism makes people happier, L = lots of people are theists. Then H&L are evidence for T iff the quotient P(H&L|T)/P(H&L|~T) > 1. We can rewrite this quotient as P(L|H&T)/P(L|H&~T) * P(H|T)/P(~H|T). Then the thread-starter’s argument at best shows that T is irrelevant to L once we know H, hence P(L|H&T)/P(L|H&~T) = 1. So in this best-case scenario, the quotient becomes P(H|T)/P(H|~T). If theists can show this is greater than 1, then H&L still ends up as evidence for theism. So that’s what you really have to be asking.
The point is that if religion does make people happier, this makes it more probable than before on naturalism that lots of people would be theists, hence weakens the evidence for theism that comes from the surprising-ness of lots of people being theists. In other words, theism’s making people happier helps screen off the truth of theism from the phenomenon of widespread religious belief.
Of course, to be evidence against theism, the happiness thing has to lower the probability of theism on all of the evidence, not just lower it on one portion of the evidence. Some theists may want to argue that theists being happier is itself more likely on theism than on non-theism, and this isn’t terribly implausible. Without having a healthier sense of the conditional probabilities involved than I actually do, I don’t know how to evaluate the overall effect on theism’s posterior probability.
Actually, I should be more clear. Let T = theism, H = theism makes people happier, L = lots of people are theists. Then H&L are evidence for T iff the quotient P(H&L|T)/P(H&L|~T) > 1. We can rewrite this quotient as P(L|H&T)/P(L|H&~T) * P(H|T)/P(~H|T). Then the thread-starter’s argument at best shows that T is irrelevant to L once we know H, hence P(L|H&T)/P(L|H&~T) = 1. So in this best-case scenario, the quotient becomes P(H|T)/P(H|~T). If theists can show this is greater than 1, then H&L still ends up as evidence for theism. So that’s what you really have to be asking.