[EM] Finding SociallyBest. Is it impossible?

[Paul Kislanko]

Peter de Blanc wrote
>>
> Perhaps some artificial inteligence tool, like neural networks or genetic 
> algorithms, or a combination of both, could be used to search SociallyBest

> (zero BR), or at least get near it. If such formula is found, it could be 
> truly complex or iloggical, something like a "black box" voting method, 
> but mathematically very good. <<

Actually, Isaac Asimov speculated along these lines back in his early 1950s
short story "Franchise", where the super-computer Multivac selected one
voter, asked him a lot of seemingly irrelevant questions, and from the
answers determined how everyone else would vote in every election. 

Pure fancy, but undoubtedly motivated by the then-new "science" of opinion
polling.

>>Zero BR is impossible with strategic voters; that would mean electing the 
candidate that maximizes aggregate utility. But if that's what you're doing,

then voters will be motivated to lie about their utility functions. It 
doesn't matter what sort of contortions you use in designing the method.<<

In the short story, and surely in practice, the AI engine knows enough about
the voter that it would account for the voter's tendency to lie. The voter
isn't asked what their utility function is (good thing, I haven't gotten
anyone to give me a reasonable definition of "utility function" except as an
abstraction for the complicated way I prioritize issues and my comprehension
of the candidate's positions on the ones that are important to me) - the
voter is asked questions designed to "profile" the voter. Based on those,
the AI engine decides what the voter's utility function is.

>>With honest, perfectly introspective voters, you could just ask everyone
to 
report their utility functions and sum them up. But such voters are a 
fantasy.<<

Utility functions are a fantasy. But it's true that asking voters to define
theirs wouldn't work. Can you define yours? I can't even get a good
definition for what that means.

>>The difficulty with evolving a voting method is that you don't know what 
strategic voting would look like. Maybe you could evolve the voting 
strategies too, but I expect you'd have pretty major issues with local 
optima.<<

This is only true if you're basing the hypothetical AI engine on the
principle of summing individual utilities. The meta-considerations I
mentioned above include that you have to start with some axioms as
underpinnings.

It might turn out that minimizing BR isn't the same as determining what is
"Socially Best".