[EM] democratic aggregation of utility

[Bart Ingles]

Forest Simmons wrote:
> 
> All else being equal, which is better, a candidate that can guarantee 55%
> of the population fabulous riches at the expense of the other 45% of the
> population living in abject poverty, or the candidate who can guarantee
> that the entire population will have at least a living wage (all else
> being equal)?
> 
> I think the second candidate would get my vote as the one with the best
> aggregate social utility plan. Note well that this second candidate would
> be the non-majority candidate in a country with a level of altruism
> slightly below our own ;-)

Unfortunately, it probably _has_ come down to "who will promise me the
best guaranteed income."  So much for self-determination.

     "One fire drives out one fire; one nail, one nail;
     Rights by rights falter, strengths by strengths do fail."
-- Shakespeare; Coriolanus 4.7


> So how might one measure this kind of aggregation?
> 
> Simple average won't do the job, because the riches of one billionaire
> compensate for the zero income of an hundred thousand peasants in the mean
> living wage calculation, without in any way guaranteeing that the actual
> transfer of funds will take place.
>
> The median is more democratic than the mean, but one above median
> billionaire still cancels one abject poverty peasant in the calculation
> without any actual transfer of funds.

This seems similar to the problem of converting dollars to utilities in
a risk situation -- i.e. which is worth more, a 50% chance at one
million dollars, or a 100% chance at five hundred thousand?  And how
much more -- would you accept a sure $400K over the 50%/$1M lottery? 
What if you needed surgery costing $900K -- the sure $400K would be
worthless in that case.

The answer is that the voters themselves should decide, not some
mechanism.


> Among well known voting methods the Approval count is the most democratic.
> In the approval count nobody's excess riches or utility is assumed to
> compensate for anybody else's deficiency.

One reason it's democratic is that it doesn't collect data only to
overrule it.  This is a problem with any so-called "hi-res" method --
ballots containing ratings can be interpreted in different way, such as
averages or medians.  And ultimately, those ratings can be converted to
rankings, and evaluated using Condorcet, or Borda, or Runoff.  And
Condorcet supporters will then point to a non-Condorcet ratings winner
and say "but this overrules the will of a majority!" and so on...


> Until we find an even more democratic (yet simple) method of aggregating
> individual utility into social utility, I suggest that we stick with
> various ways of implementing Approval.

Various ways?  I suppose you can come up with other methods, and include
"approval" in whatever you decide to name them, but that doesn't make
them "approval voting".  And then you can spend the next 20 years
performing various logical and mathematical analyses, trying to rule out
any fatal flaws.  And then try to recruit some credible organizations to
demonstrate the method for a dozen years or so, before attempting to
sell it as a public proposal.  With any luck, this will come about the
time the public is ready to abandon IRV.

> In particular, we can implement Approval in various ways that overcome the
> "lack of expressivity" objection of its opponents, and without having to
> apologize if Approval doesn't happen to pick the candidate with the
> greatest mean or median rank or rate, or any other candidate that would
> have won under some other inferior standard of democracy.

If you collect either rankings or ratings, you will either have to
apologize for not choosing the greatest mean or the greatest median,
since it is impossible to guarantee both, let alone satisfy every other
possible standard.

I think the "lack of expressivity" is largely imaginary, or else trumped
up by IRV advocates.  The best way to counter the latter is to nuke the
pro-IRV arguments; for the former, try education.  Sure it's possible to
come up with examples where a voter is ambivalent toward all permissible
strategies, but try finding one from a real-world example, such as a
real election from the last 20 years.

Besides, if you want expressivity, get a bumper sticker.  Or a lot of
bumper stickers -- possibly arranged in order from favorite to least
favorite...    :-)

Which gives me another idea -- why not do the same with campaign
contributions?  You could rate your candidates from 0 to 10, and then
donate in proportion to your rating.  So if you have a 10, 6, 4, and 0,
and you have $1000 to contribute, you could donate $500, $300, and $200
to the first three candidates.  Might not be the best use of a limited
resource, but it would be expressive.

Hanging it up while I can still see,
Bart


> Use some kind of ranking or rating ballot (like my five slot grade
> ballot), and allow the voter to mark the Approval cutoff (directly or by
> virtual Minimum Acceptable Candidate) or allow (by check box) Cranor's
> optimization to choose the Approval cutoff for that ballot.
> 
> Peace to All,
> 
> Forest