[TrainedMonkey]

1. To the extremes that environment of the problem will be able to support. For example if you want to figure out education one of the extremes would be one teacher per student, other extreme would be the top minds in the field create a single class online and teach it to everyone together (Everyone gets same material, as opposed to different material per student in first case). You immediately see what is different between two, second one enables way to collaboration and idea swapping between students at the cost of face to face to face teaching and individual attention each student receives. In this context you can think of how many students can each of the extremes support.

2. I think analogies are a way of simplifying problem so that you can model it. People usually do it by assigning actors that they have experience with to the models, but that is not necessary. So if we model education as knowledge transfer between actors we can look for other concepts that share similar models.

3. Bad examples are a problem, I have nothing on this one. It might be mitigated by the fact, that on average some people will get good examples and succeed, but that feels like a non-argument.

4. Good point on historical factors, they are still useful to help pinpoint what kind of factors are affecting problem you are thinking about.

5. I think point of thought experiments is to try to come up with an obvious flaw in your experimental setup. Knowing those potential flaws would help you to set up controls in real experiment.

Not everyone is going to succeed in understanding big problems such as education system anywhere close to accurate. Even fewer people will be able to do anything with their understanding. I do not think that is the problem, as long as people try some will make it. I do know for sure that if you do not try, you will definitely never make it.