Isn't this tautological or trivial because any proposed alternative requirement than reward maximisation can be reframed as maximising a different reward?