Hedonistic Utilitarianism—produce the most pleasure.
Actual (not necessarily revealed) Preferences
Ideal preferences—produce the most of what people want to want, or would want under ideal reflective circumstances
Welfare Utilitarianism—produce the most welfare, which may differ from preferences if people don’t want what’s best for them.
Ideal Utilitarianism—outcomes can have value regardless of our attitude towards them
In every kind of utilitarianism, including revealed preference, utilities are assigned to outcomes. The varieties I’ve described, and revealed preference, just disagree about how to assign values to outcomes.
These are different mechanisms to theoretically quantitate utility, but do they actually have implementations? (Revealed preference is unique in that it’s an implementation, although a post-hoc one, and defined by the fact that the utility is qualitative rather than quantitative—that is, utility relationships are strictly relative)
None of these actually assign utility to outcomes, they just tell you what an implementation should look like.
I’m not sure what you mean by an implementation if you think revealed preference is an implementation. We don’t have revealed preference maximising robots.
Hedonistic Utilitarianism—produce the most pleasure.
Actual (not necessarily revealed) Preferences
Ideal preferences—produce the most of what people want to want, or would want under ideal reflective circumstances
Welfare Utilitarianism—produce the most welfare, which may differ from preferences if people don’t want what’s best for them.
Ideal Utilitarianism—outcomes can have value regardless of our attitude towards them
In every kind of utilitarianism, including revealed preference, utilities are assigned to outcomes. The varieties I’ve described, and revealed preference, just disagree about how to assign values to outcomes.
These are different mechanisms to theoretically quantitate utility, but do they actually have implementations? (Revealed preference is unique in that it’s an implementation, although a post-hoc one, and defined by the fact that the utility is qualitative rather than quantitative—that is, utility relationships are strictly relative)
None of these actually assign utility to outcomes, they just tell you what an implementation should look like.
I’m not sure what you mean by an implementation if you think revealed preference is an implementation. We don’t have revealed preference maximising robots.