Informed, self-regarding preference responds to prospects, but not directly to number of exact copies when above one instance.

Does it matter to me that my exact experience is multiply instantiated across the multiverse? Not really. Does it matter to me how my future is going to go? Yes, it matters a lot to me. And both answers hold for all my identical copies, who from the inside face the same situation and the same distribution of futures, as I do.

It seems to me that after informed reflection, people might happily trade off their number of exact copies / measure for higher welfare if given the opportunity. It affects only them and their exact copies, who share their preferences and felt welfare and don't feel changes in copy count, and it affects no one else.

Yet total utilitarianism would be against it if it "reduces total welfare". Against the type of experience's interests, total utilitarianism can recommend a) concentrating suffering into one exact copy, b) diluting extreme happiness into many exact copies. It seems like something's up with total utilitarianism's utility function.

If you think:

1. Welfare of observer-moments should be considered on a type level rather than a token level (i.e. since from the inside it's all the same to you how many copies you are), and that
2. You should account for which observer-moments they become next (i.e. subjective continuity is worth caring about)^[1]

Then you directly end up with a Markov reward process over types of observer-moments. The sum of value functions (or a well-defined analogue in the non-terminating case) works as an impartial utility function over the different types.^[2]

In large worlds where all types of observer-moments exist, you can't change the state space / set of all types of observer-moments - but you can change the transition probabilities from observer-moment to observer-moment.^[3] So, this makes the "flow ethics" here action-guiding in large worlds.

Working on a longer writeup.

1. ^{^}

   I still do care about my future even if we might debate about whether subjective continuity is real.

2. ^{^}

   Impartiality means counting every type of observer-moment's future equally, so later moments appear more in the sum by being part of more futures. This seems fair even if the math looks jarring.

3. ^{^}

   This isn't a claim that physical measure doesn't matter. Because it does matter - but it's involved in a different way to how measure-weighted total utilitarianism uses it. Measure-weighted total utilitarianism uses measure to set moral weights. Here, measure is involved only via setting transition probabilities. Transition probabilities seem most naturally the ratio between measure of a particular successor of a predecessor, and the measure of all possible successors of that same predecessor.