What causes flutter?

An aerodynamic force pushes or displaces the surface. The surface then moves until it's own stiffness or it's aero shape halts the displacement. At that point, aero force usually contributes to "replace" the surface where it came from. The structure of the surface contributes to the return as well. During this displacement, the mass of the surface contributes to the forces involved. The greater the mass of the surface, the greater the contribution.

OK, the surface has been displaced, it has moved as far as it can, and for several reasons it's starting to return "to center". The mass of it really contributes now. It aids the force and speed of the return motion. So the surface now passes center and repeats the deflection in the opposite direction, but with more energy than at the start. etc etc etc ....... until you either get off the throttle and the model slows down, or POW!!!!

Flutter is an oscillation that is self strengthening (usually). Rodney's post covers it pretty well. And he clarifies the contribution of mass balancing better. But one detail about mass balance is worth mentioning.

Our models are often fairly close to flutter. And sometimes it only takes a couple mph more speed to get flutter. And sometimes our rigging strength or stiffness needs just a bit more strain to fail. Have you noticed that flutter often happens on a pullout? It also happens on really hard turns. Centrifugal force is real and is a significant contributor to the load our rigging sees. When a surface is not mass balanced, the entire weight of that surface bears down on the rigging. Look at the surfaces on an IMAC model when it's parked. Gravity pulls all those surfaces and they droop. Mass balance them and they're less apt to droop. But the purpose of mass balancing isn't just to have the airplane look better when parked. When that sucker is pulling it's guts out around a turn or screaming out of a dive, those surfaces are pitting more than their own weight against the rigging and the servo(s). You know, models often see flutter when the load on the surface is more than the servo can handle. And constantly overstressing the rigging wallows out the fittings. But hang a mass balance across the hingeline from that surface and what happens? The mass balance sees the same force the surface sees but works to lessen the force the rigging sees. It cuts down on the wear and tear on the rigging and the load on the servos. When it helps and what it helps is sometimes it's importance.