Armchair Science
aeon.co
aeon.co
However the cannonball experiment still needs to be done because it's actually two parts: you can't say what will fall at what speed, the cannonball or the musketball, due to the clever argument that connecting them creates a contradiction. But you also made the prediction that they fall at the same rate, which you can try out.
In the end it seems thought experiments are basically what we do in intro physics classes: they're used to decide what we actually predict will happen. It still takes a bit of grinding to get a number out of a model, and a lot of effort is used in understanding why the model predicts whatever it does, and when the model stops being applicable. Only rarely do we actually test against reality.
1. A prediction or thesis which then then prompts a theory and then experiments to verify. The EPR paradox is an example, the cannonball is another.
2. The are an argument to convince someone/explain something in a physical discussion. The text given here is an example.
Now one could argue that someone learns something in the second case (and that someone could even be everyone), but technically we already know everything. It's sort of like we already have a (experimentally) verified theory and are just investigating some corner case. Or otherwise we end up at option 1, where we need an actual experiment to probe what the world is really doing.
To me, thought experiment always comes first. Sometimes the result is obvious, and sometimes it's not. The most exciting ones are when you're not sure what will happen and can do smaller experiments to refine your model over time to get closer and closer to the (possible)truth.
Thought experiments are particularly exciting to me because of the ability to scale time or space in different ways. You can watch e/m waves interact with electrons, or watch a canyon erode over millions of years.
Man I love science.