I think you're right about the magnetic fields. I'm pretty sure the interactions between the metronomes in the video can be modelled as message passing between nodes in a square lattice of variables, with stronger association potentials in the direction of the swing than in the direction perpendicular to it.

As far as I understand it, this is similar to the Ising model of ferro-magnetism [1], where magnetism arises from the interactions of neighbouring atoms.

For more on this kind of stuff, look into graphical models. Coursera has a Stanford course with Daphne Koller on it.

1: http://en.wikipedia.org/wiki/Ising_model