In my last post I pointed out that evidence gleaned from experiments must be used to arbitrate between the possible explanations of an observation. But what constitutes evidence?
A proper experiment changes one variable in the system, and valid evidence is the observation of the final state of the system after an experiment is performed. If there is no change between the initial and final states, that counts as evidence against a particular hypothesis. If there is a change in an expected manner, that counts as evidence for a particular hypothesis.
In the car example that I used before, we have a number of hypotheses for why the car won’t start: it may be the alternator, the spark plugs, or the battery. We replace the alternator and the car still won’t start. The initial and final states are unchanged. That’s evidence against the alternator hypothesis. We replace the spark plugs and the car won’t start. The spark plug hypothesis is ruled out. We replace the battery and the car starts. There was a change between the initial and final states of the system. We have evidence that confirms the battery hypothesis.
But what kind of experiment can we perform to test the God hypothesis? In each of the other experiments, we had to physically interact with something in order to confirm or disconfirm a hypothesis. That physical interaction was the causal connection between the object of the hypothesis and the state of the system. How do we interact with God, or with something else, to make a causal link between God and the system? What do we interact with to perform the experiment?
We are stuck at the starting line. Judging by the variety of religions and spiritual traditions in the world, there is considerable confusion over what the nature of God is and how to interact with Him. God is a poorly defined concept because there are so many definitions. We have a clear and concise concept of a battery, which means we have a clear way to test the battery hypothesis. We don’t have this luxury with the God hypothesis.
There’s a crucial point to be made here. Evidence is a hypothesis-killer. Without evidence, you can invent hypotheses endlessly, but once you acquire evidence, you start ruling them out. It is the only way to arbitrate among them. We have a clear definition of a battery because we have evidence of batteries. We can observe them. We can build them. But we have innumerable concepts of / hypotheses about God precisely because we lack evidence to arbitrate between them.
If you understand Bayesian reasoning, you understand why this does not bode well for the God hypothesis. Right now the God hypothesis is hopelessly muddled and lacks any means of testability, and this situation doesn’t look like it’s going to be resolved soon. Therefore, we are justified in rejecting the God hypothesis, at least for now.
 The change must be of an expected manner. If we change the battery and the car explodes, there was a change between the initial and final states, but we can’t conclude that the battery was the cause of the car’s inability to start.