Shared publicly  - 
Okay, so scratch the idea of maximising total human happiness. How about maximising root-mean-square human happiness? Engineering is all about knowing the best kludges.
Tom Anderson's profile photoGiles Lewis's profile photo
Pffft I don't think much of a social welfare function that doesn't satisfy Pareto optimality.
Both sum and RMS assign positive utility to Pareto improvements, and negative utility to Pareto, er, disimprovements. They may also assign nonzero utility to changes which are neither. But that's always going to be the case in a useful welfare function.

Basically, Pareto optimality is wack.
Felix would also break root-mean-square happiness eventually, though. Indeed, once everyone else's happiness went negative it would go even more wrong. Whereas he's not allowed to start eating people's bread under a pareto regime.
Can you have negative happiness? If so, then RMS is definitely useless.

Under a pareto scheme, Felix can't eat everyone's bread, but you also can't tax Fred Goodwin to buy everyone bread.
I don't think that sort of analysis usually includes negative happiness, but the SMBC strip did. (It was the strip I was referring to in my first comment, anyhow.)

Under Pareto schemes, you can tax Fred Goodwin, you just have to compensate him for the utility loss somehow. If people are taken to have diminishing marginal utility from income you can do that and still make others better off, leading to a net gain (in both total utility and Pareto efficiency terms). The problem with Felix is that he has increasing marginal utility, so the two criteria contradict, and apparently SMBC's imagined computer decided to resolve this by dumping the Pareto one. So I thought it was a bit of a straw man. Straw computer. Whatever.
Add a comment...