Mantle convection is usually treated as a far-from-equilibrium self-organized system (SOFFE). Ilya Prigogine used thermal convection as the poster-child of self-organization. But if the top of the fluid contains plates, continents and sources of dissipation – plate bending, cracking, buckling and folding – to replace viscosity, then the top, rather than the interior, may be the self-organized system and may control and organize mantle convection.
Very few mantle convection simulations allow for the necessary degrees of freedom, allowing the top and the plates to self-organize, and the self-organizing principle is not known (is it maximum dissipation, maximum entropy production, or what?). Consider also Marangoni convection, a related, top-down form of convection, that mimics thermal convection, e.g. Rayleigh-Benard convection.
This is a fertile field for research but has so far gone largely untapped in the mantle convection and geodynamic communities.
Don L. Anderson