The old ideas: (1) The "total view", rather than the "prior existence view." We must take into account the total result of our choices, not limit our focus to the impact on those who exist prior to, or apart from, our decisions. Both views lead to problems, but the second to worse problems. (2) Preference utilitarianism, not hedonistic utilitarianism. The good is desire-satisfaction, not simply happiness. People have vastly more desires than sheep, so potentially more satisfied desires, so potentially more good in their lives.
Take the total view plus preference utilitarianism together. Which universe should we aim for, based on that package? It depends how we evaluate preference satisfaction. We could take every chunk of preference satisfaction as a positive good. In that case, the Peopled Universe is best, because there's the most preference satisfaction in it. There's none in the empty universe, and less in the Happy Sheep Universe, because of the more limited potential of sheep to have preferences.
But that way of evaluating preference satisfaction is problematic. If I make you have a burning desire to eat marshmallows, through some devious chicanery or other, do we really want to say it's positively good if you get to have lots of marshmallows? An alternative is the debit view of preferences. We take an unsatisfied preference as a debit in a sort of ledger. When the preference is satisfied, the debit is erased. Making you want marshmallows is actually bad, but I can cancel it out by feeding you marshmallows.
With this adjustment, we now get a different ranking of the three worlds. The Non-Sentient Universe comes in as #1, since the others have uncanceled debits. This sits unwell with Singer (and with me, too). There are other ways you could look at preference satisfaction, having to do with the way preferences are generated, but Singer makes a much bigger move away from 2e. He suggests (p. 117) that what has value is not just preference-satisfaction--
We could try to distinguish two kinds of value: preference dependent value, which depends on the existence of beings with preferences and is tied to the preferences of those specific beings, and value that is independent of preferences. When we say that the Peopled Universe is better than the nonsentient Universe, we are referring to value that is independent of preferences.Like what? What, besides a satisfied preference, might have value?
We could hold a pluralist view of value and consider that love, friendship, knowledge and the appreciation of beauty, as well as pleasure or happiness, are all of value.The Peopled Universe has more of those positive goods than the Non-Sentient or Happy Sheep universes, so it's best. Singer's tone in these passages is exploratory and respectful of the difficulty of the issues and the diversity of positions. Benatar's preference for the empty universe gets a respectful nod. But he seems willing to complicate his own ethical theory in order to avoid the ultra-gloomy view that the best world is the most barren.