For context, in classical mechanics we also have to perturb the system to measure it, but we can make the perturbations arbitrarily small. That's why we call it

For context, in classical mechanics we also have to perturb the system to measure it, but we can make the perturbations arbitrarily small. That's why we call it a measurement rather than an interaction, and can get away with not explicitly modeling it. A seminal paper on the question in classical mechanics was the Bohr Rosenfeld 1933 Paper on measuring Electrodynamic fields, that showed that classical EM fields can indeed be measured accurately (I can't find it online right now, but the follow up on QED is [1]). The question appears in the context of General Relativity as the question when extended bodies follow geodesics. These are all different to the situation in QM. There is an irreducible alteration of the state in QM. It resembles in many ways the updating of a probability distribution if you gain new information. Imagine you have two envelopes with a blue and a red card in them. You mix them and send one of them far away. Now you know that the far away envelope has a 50/50 chance of being blue. It is in a superposition of red and blue. If you open your envelope and look in, the superposition collapses, and you suddenly know exactly what state the other envelope is in, even though you didn't touch it, didn't disturb it at all. QM measurements behave just like this "updating of probabilities upon learning something" _except_ that we know from other, more subtle properties (Bell's inequalities) that they are _not_ probabilities referring to a hidden state. [1] https://journals.aps.org/pr/pdf/10.1103/PhysRev.78.794 https://journals.aps.org/pr/pdf/10.1103/PhysRev.78.794