Because if your choices under uncertainty do not maximize the expected value of some utility function, you are behaving inconsistently (in a particular sense, axiomatized by Savage, and others - there's a decent introduction here).

These axioms are contestable, but the reasons for contesting them have little to do with population ethics.

Also, as I've said before, the utility function that consistent agents maximize the expectation of need not be identical with an experienced utility function, though it will usually need to be a positive monotonic transform of one.