The main idea is that we should analyze evidence, confirmation and testing by focusing on procedures. That procedures for collecting samples are relevant when applying statistical methods like Bayesianism. For example, in the the-Raven’s-problem, a possible solution is to pay attention to how the observations of ravens are being collected.
Another idea is that in experiments, interventions provide better evidence about cause and effect than passive observations. This follows a similar thinking to Hacking-on-scientific-realism.