348 RELIGION, MORALITY, FAITH, AND REASONbeliefs are necessarily not unbreakable beliefs. There is not only no point
in trying to make them so; the hypothesis that they are is simply self-
contradictory. Nor is there anything to lament here; it is not sensible to
lament that the logically impossible does not obtain. There is no loss here –
nothing that could not possibly be so amounts to a loss of anything. That
religious beliefs are not unbreakable beliefs should be kept in mind as we
consider faith and reason. It will help us do this realistically.^4
In order to have a neutral expression, let us say that if Kim believes that
her golden retriever wants a treat – remembering that Kim’s golden
retriever wants a treat is neither a necessary truth nor something whose
truth follows from Kim’s believing it – let us say that her belief is a piece of
delicate knowledge and the belief that constitutes it a delicate belief. A
delicate belief can be true, well supported by evidence, and reliable. A
delicate belief can be false, against the evidence, and unreliable. Much of
what we know is constituted by our true delicate beliefs.^5 Delicate
knowledge is not a defective version of unbreakable knowledge; it is not
any kind or version thereof.^6
Probability
Delicate beliefs are typically, though perhaps misleadingly, said to be only
probable. What is meant by this is sometimes clear and sometimes unclear.
The proposition The odds of getting a six with a fair throw of a fair die in a
fair environment is one in six is a necessary truth. But the beliefs that the
die, or the throw, or the environment in question are fair are delicate
beliefs, and if they have any probability, the proposition ascribing that
probability to them is not a necessary truth.
Suppose that it is a probabilistic law that Given the occurrence of an
event of kind A, the chances of the occurrence of an event of kind B is 99.9.
Assuming we know that this is a law and that an event of type A has
occurred, we know that the odds are 999 to 1 that an event of type B will
occur. Here, we can quantify our probabilities. We can also quantify our
probabilities in the absence of any known laws. Suppose that we have 99
marbles in a bag, 33 each of blue, red, and green. Assume that conditions
are such that our odds of drawing a red marble from the bag is one in three
(being blindfolded, we cannot look into the bag and pick what color marble
we want, etc.). A reliable friend tells us that the marble we take from the
bag is not green. So we know it is either red or blue. There are now 98
marbles in the bag, and 33 of them are green. Either there are 33 red and 32
blue, or 32 blue and 33 red; we do not know which. So now we can infer
that, relative to our knowledge, the odds of our picking a red marble are