Science is inherently atheistic

[Angrboda]

The starting point is always "null" or "no relationship", and then we try to determine the probability of relationship by controlling things that would skew those results, that way we are only testing the specific relationship in the research between two or more subjects.  If we find nothing, we conclude the null was correct, but we also give a numerical value on the probability that "null" was the right result. 

Just a minor point, no, we do not conclude that the null is correct, only that the null is not necessarily incorrect. Those are not the same thing, though they are often confused.