It behaves this way because that's how it was built. By adjusting the mass distribution, we could make a scale that flops to one side, is roughly balanced at all angles, etc. However, those scales would not be useful, so the scale isn't built that way.
It might be assumed from the left/right symmetry of the picture that the system cannot decide which way to go, and so is at an equilibrium point. This equilibrium will be stable if a small perturbation (rotating the beam a small angle) raises the center of mass. It will be unstable if a small perturbation lowers the center of mass.
Beyond that, it is difficult to say how the center of mass moves simply by looking at your picture because we do not completely understand the mass distribution and the location of the pivot point.
When finding the center of mass, we can ignore any stationary pieces because we are only interested in the change of the height of the center of mass. Additionally, if the pans hang freely down, it appears as if one will rise by the same amount the other falls, and thus they will not change the height of their center of mass when considered jointly. They can also be ignored.
Let's assume the rest of the scale rotates rigidly. In that case, the center of mass of the rigid portion we're considering will be constrained to a circle with its center at the pivot point. If the center of mass is exactly at the bottom of the circle, we have a stable equilibrium. Otherwise, it is unstable.