It seems to me that the most defective assumption is that the well being of the whole is some linear combination of individual preferences, up to very large numbers, and in very atypical circumstances (e.g. involving copies).

You can't expect that you could divide the universe into arbitrarily small cubes (10cm? 1cm? 1mm? 1nm?) and then measure each cube's preferences or quality of life and sum those.

So, on the purely logical grounds, summing can not be the general rule that you can apply to all sorts of living things, including the living thing that is 3cm by 3cm by 3cm cube within your head.

I thus don't see any good reason to keep privileging this assumption.

If you are a philosopher, and you want your paper to look scientific, then you need mathematical symbols and you need to do some algebra so it looks like something worthwhile is going on. In which case, by all means, go on and assume summation, this will help write a paper.

But if you are interested in studying actual human ethics, it is clear that we evaluate the whole in non-linear ways - we have to do that to as much as recognize an ethical value of a human despite not recognizing an ethical value of a quark - we don't think that any individual quarks within a human feel pain, but we think that the whole does.

Conversely, for a very large number of pedophiles, we do recognize that an individual pedophile wants to watch child porn, but we do not sacrifice a child, however large the number of pedos is. And it is not any more incoherent than recognizing that an individual person has preferences but the elementary particles do not.