In Austrian economics using the framework of Praxiology the claim is made that preferences (the rough equivalent of utilities) cannot be mapped to cardinal values but different states of the world are still well ordered by an individual's preferences such that one world state can be said to be more or less desirable than another world state. This makes it impossible to numerically compare the preferences of two individuals except through the pricing/exchange mechanism of economics. E.g. would 1 billion happy people exchange their own death for the existence of 1 billion and 1 new happy people? To answer the question simply ask them and what they would do or observe what they do in that situation and that will reveal their preferences.

Taking a preference-based approach, consider the set of all individuals and the set of all world-states. Each individual has a well-ordered list of preferences of possible world states, with the only restriction that it be bounded from above by a maximal preference. In every world state the next world state is chosen by all individuals voting for the most preferable next reachable world state. In a majority voting system each individual votes for its maximally-preferred world state. In runoff and approval voting the first, second, third, etc. choices are the highest, next-highest, etc. ranked preferences for world states, respectively. Ethics thus reduces to the problem of fair voting.

An obvious criticism of Austrian economics is that it simply describes the economy as the result of all individual actions (with which individuals reveal their true preferences, by definition) with no additional predictive power. I think that by contrasting the theoretical results of perfect ethical preference voting with the theoretical results of perfectly calculating a utilitarian theory, there may be some insight. The basic difference is that economics relies on pricing to build an economy but in ethics we can cheat and ask theoretical questions about all possible world states.

Potential or hypothetical individuals would have their own preferences for world-states but, as the article mentions, their preferences may not be compatible with the set of possible next world states that we are voting on If those hypothetical individuals never have enough votes for possible next world states then they will never have any influence. Individuals currently sleeping, anesthetized, or frozen in liquid nitrogen have hypothetical preferences for future world states that may very well coincide with our preferences for future world states, and therefore they have a greater chance of existing as acting individuals in our future world. Ultimately in any ethical theory it is only our estimation of a hypothetical being's preferences that we can consider, so their preferences are subsumed into our own preferences.

3^^^3 people will probably rank their preferences for world states with and without a single dust speck in their eye as nearly indistinguishable, but world states with torture are hopefully quite lower in rank than equivalent world states without torture. The one person whose torture depends on the vote may prefer world states with 3^^^3 dust specks far more than world states with 50 years of torture, but their vote clearly doesn't matter in any conceivable voting system. Nevertheless, so long as the existence of torture is more repugnant than a single dust speck, 3^^^3 people will vote to receive the dust speck instead of allow that individual to be tortured.

Populations of nearly any size will probably not vote to replace themselves with a different population (whether of humans, paperclips, or smily-faces).

There are still problems: Bacteria and parasites may deserve a vote. Weighting may fix that problem. Hated minorities are still at a disadvantage even in the fairest voting systems. On one hand if people are not personally inconvenienced by the actions of a hated minority they will probably prefer worlds where that minority is not tortured over worlds where they are tortured, simply because of their general aversion to torture. On the other hand a large number of voters in democratic countries have not kicked torturers out of political office. This is distressing because far fewer than 3^^^3 people have been affected by, say, Maher Arar. There is apparently a tendency in humans to have a preference for the brutal punishment of an assumed criminal even if there is only a tiny marginal chance of value to themselves. I think this is a failure of rationality and probably not a failure of any particular ethical system.

The primary difference between additive functions of individual utility and preference voting is that the most important effects on individuals have the largest influence on their preferences. There is no correspondence to having a maximal utility for not having a speck in one's eye and also a maximal utility for torturing another individual. One or the other will have strictly greater preference. In approval or runoff voting the preference for torturing another individual will fall behind a series of other more pleasant preferences unless that individual actually had a major effect on the voter. In effect everyone is forced to vote for what really matters to them instead of arbitrarily ruining another individual's life for no appreciable benefit. Utilitarianism could conceivably have exactly the same ranking of values as preference voting (just enumerate all the N preferences and assign them values of i/N from least to most preferred) but there is no guarantee that an individual faced with choosing utilities would assign the same relative value to world states as an individual choosing preferences.

It appears that Eliezer went down this road a ways in http://lesswrong.com/lw/rx/is_morality_preference/ and then went off in another direction before enumerating the idea of what would happen if everyone voted based on their preferences and then acted to achieve the winning world state instead of acting to achieve only their own maximally preferred world state.