>>12678918This exactly. Testing a single hypothesis will get you nowhere. Rejecting the "null hypothesis" tells you almost nothing about how probable your testing hypothesis is given the evidence. The probability of a hypothesis is meaningless in the absence of another hypothesis to compare it to.
To add more details to the mentioned steps:
1. Generate as many hypotheses as you can, each compatible with current knowledge.
2. Weigh the probability of each hypothesis by its "simplicity" in the Occam sense.
3. Come up with an experiment and assign a probability distribution to the possible outcomes for each hypothesis.
4. Run the experiment, Given the outcome, you get a ratio of likelihood between all hypothesis depending on how much probability each assigned to the outcome. You can use this to update your probabilities for each hypothesis.
5. Repeat.
