What Causes a Measurement to Change the State of a System in Quantum Mechanics?

[Kara386]

I read that measurement changes things because we have to bounce photons off an object to 'see' it and that changes its position, momentum etc...

But on the other hand, Griffiths' QM book seems to suggest we don't know what it is about measurement that changes the state of something. We don't know what's special about measurement, or what exactly constitutes measurement.

The photon idea sort of makes sense to me, so if it isn't actually the accepted answer, then why not? It's just that the photon thing and Griffiths seem to say different things.

From quickly googling this, most people seem to say that the act of measurement will always disturb the thing being measured. That doesn't seem particularly strange, does it? But until now, having only read books for laymen, I'd got the impression that the fact measurement changed things was bizarre and no-one knows why it changes things or even what exactly constitutes measurement... I'm quite confused.

 

[Simon Bridge]

A measurement has occurred when the state of a system has an effect outside the system... broadly: the measurement apparatus.
Griffiths is broadly correct - the situation that is mysterious is the process by which a state turns into another state. It is often called "decoherence" or "wave function collapse". We do usually know the physics behind the devices we use to make a measurement or we would not be able to build them reliably.
Perhaps look up "Stern Gerlach experiment".

The photon approach is not accepted because, short answer, it is wrong: it gives the wrong idea about the statistics.

Note: a measurement of a system does not have to disturb the system.
For instance, measuring the energy of a system prepared in an eigenstate of energy will not alter the energy state of the system.
Usually the "measurement disturbs the system" comes up in context of Heisenberg's uncertainty ... where measurements of position affect the result of subsequent measurements of momentum. But there are other things we can measure: it is possible to prepare a system in simultaneous eigenstates of two or more measurable things ... in which case a measurement of anyone of them will have no effect on the outcome of measurements of any of the others.

Nobody feels that a measurement changing things is bizarre at all - of course interacting with a system can change it: if I poke you, it would be worrying if you didn't react.

The pop-sci books and videos are trying to invoke some sense of spookiness with QM and they don't usually do a good job.
Your best bet, if you want to pursue consequences more deeply, is to start an introductory course in QM - needs maths.
Otherwise, just ignore attempts by pop-sci media to make you think anything mysterious is happening.