> The part that the scientific method lacks is determining what a good implementation looks like.

That seems like a feature, not a bug, to me. Descriptions of good implementations are likely subjective and vary based on what is being tested. It may even change over time as human knowledge improves.