observation → induction → hypothesis → test hypothesis by experiment → proof/disproof → knowledge
The basic tenet is that we can test ideas. We want to know how something works and we do repeated experiments and others can do the same experiments. This limits our tests to the present.
To get around this limitation the term science has been expanded. People make hypotheses about single events. Events that by their nature are not repeatable. The problem is that because the term "science" is used for this also people equivocate without being aware of it and in philosophical debates conflate the meanings. This leads to absurd accusations such as creationists investigate gravitation by reading the Bible whereas evolutionists drop bricks off buildings.
What is the difference?
- Repeatable science is called operational or empirical science.
- Investigating events is called inferential or historical science.
Proving that an event can occur does not prove that it did occur (though it may add evidence to the proposal). Showing that cats can be killed by drowning does not prove that the dead wet cat found in the rubbish bin was drowned. Proving that an event cannot occur does however counter wrong theories. Showing that jelly cannot cut skin proves the man was not stabbed to death with a giant pudding.
Inferential science in forensics is presented as circumstantial evidence. This however is not the only evidence. Witness is the other type of evidence. And a trustworthy witness is worth more than inference (though the 2 can be compatible in ways that are not immediately obvious).
There is very little difference between creationists and evolutionists about operational science. It is repeatable by anyone and the results can be seen and rechecked. The theory that best explains the data (eg. classical versus relativistic physics) and the significance of a particular experiment may differ, but the difference is still minor.
The major differences in worldview are in inferential science; the grand theories that are invoked to explain history. As we cannot observe this event due to its singular nature (that is, it occurred once) we are left with inferring the likelihood of various scenarios. Data is either consistent with these theories or not, but even showing that a fish can evolve into an amphibian now does not guarantee it has previously; it would be strong evidence that the theory is correct though. Other good evidence for the veracity of a theory is predictive evidence. Because much of a theory is modelled on observations it cannot be said to be predictive; the data is known first and the theory is derivative. Further, credence of the explanatory power of a particular paradigm is overrated, there is no end of sub theories that can be thought up after the event. Coming up with predictions that are different (therefore discriminatory) for several competing theories that can subsequently be tested can add credibility to the correctly predicting theory.