I would have loved to have had a course in school about "The Design of Scientific Experiments." One that described the processes of landmark historical experiments from antiquity onward, and challenged students throughout: "Given this set of constraints, how would you design and execute an experiment to estimate the size of the Earth? Disprove phlogiston and luminiferous aether? Measure the speed of light?"
I don't think many people today would be able to propose the Michelson Morley experiment and then actually do it. It was truly heoric (and Michelson was a genius).
We did this oil/water experiment in freshman physics or chemistry lab. It was rushed, everybody just did the minimum, the teachers barely explained any of it, and then we moved on.
I agree. The Michelson Morley experiment reminds me of some difficult algorithms: simple only in hindsight, and implementation is _hard_ to do correctly.
Experiments are HARD. There is a joke among physicists that theoreticians are washed up by 35 but experimentalists don't even get started until 45.
To make a physics experiment work you have to be ridiculous about recording details and have a strong intuition. You have to design the experiment such that you can differentiate between "hypothesis wrong" and "equipment doesn't work" because you don't know the answer.
(For example: When they turned on LIGO for the first time, they almost immediately caught a great event. Huge victory party, right? Nope. They promptly ignored it assuming that something was wrong with the machine. And it was only after significant post analysis and correlation that they decided that it was a real event.)
And this is my sticking point with a lot of "Science skeptics" around that have skepticism as their personality
Make no mistake, I do take scientific discoveries and knowledge very serious, and knowing the stories make it appreciate more the efforts and the work it took to get there
But a lot of times people think the experiments give a very clear-cut results, when it's more like "one line is squiggly down and the other is squiggly up" with data being barely over 5 sigma
