In The Outsider Test for Faith, John Loftus often makes statements like:
Faith, as I argue, is an irrational leap over the probabilities. Probabilities about such a matter are all that matter. We should think exclusively in terms of them. (19)
If God created us as reasonable people, then the correct religious faith should have sufficient evidence for it, since that’s what reasonable people require. (22)
Because science is the only game in town. It works. It produces the goods. There isn’t a better alternative. (119)
Loftus' attitude seems representative among many atheists; while I'm aware that there are other views, it is this kind of view which leads to the title of this question: Ought we only form beliefs based on sufficient empirical evidence? Here are some sub-questions which may help guide answers:
How can we get an ought from an is, given that evidence only describes what is? Reformulating the title as a hypothetical imperative doesn't seem to help, because the scientific method doesn't seem to ever make a hypothetical imperative binding. And yet, surely those who would agree with the quoted text would find certain hypothetical imperatives binding?
Given that the scientific method includes "basic arithmetical truths and also certain truths about formal provability", Gödel's second incompleteness theorem seems to apply: a given axiomatic formulation of the scientific method will be unable to imply certain (virtually all, it turns out) truths. To say that science is complete—there are no other methods of knowing things—seems to imply that science is inconsistent. Is this 'seems' an illusion?
Does the scientific method require that we make observations prior to forming hypotheses? It could be argued that probabilistically, trying to come up with hypotheses before making a 'sufficient' number of observations will result in a failed hypothesis too much of the time. Is there evidence to support this which can probabilistically rule out all other kinds of hypothesis-forming?
Must we accept any unfalsifiable claims? The answer seems to be 'yes', given e.g. the example of crystal structure discussed by Michael Polanyi in Personal Knowledge (43-48, 1984 paperback). A set of geometries seemed to well-describe many crystals found in nature, and even if some crystals aren't well-described, that doesn't falsify the instances in which it does. This opens up the question of which unfalsifiable claims we ought to accept.
If we model human reasoning as Bayesain inference, we can ask what is meant by 'sufficient' by what the 'probabilities' are, prior to any observations: what universal prior ought we start with? I am tempted to say "the most effective one", but that presupposes a purpose; are we given an objective purpose?
I am left with the suspicion that one ought to only follow the scientific method to the extent that one only wants to predict one's future observations—this smells like a tautology. It seems to beg the question to say that the most important area of 'truth' is objective reality, a question which cannot be answered with the scientific method. A 'yes' answer to the main question would seem to fall prey to something like the error of Logical Positivism.
I was tempted to phrase my question as:
Is science the only source of truth?
I believe this is a bad question, for a few reasons:
- We care about whether a method can reliably arrive at truth/knowledge.
- The terms 'truth' and 'knowledge' would possibly stir a debate about scientific realism vs. instrumentalism, which may not be a prerequisite for the answer to my question.
- My question makes explicit that we are talking about what one ought to do, which is not explicit in the above formulation.