The quadratic voting people have a company

[Clay Shentrup]

http://collectivedecisionengines.com/

Try not to giggle as you read this excerpt:

Quadratic Voting is a way for groups to make decisions together that achieves the greatest possible good for the greatest number of group members. 

It was invented by University of Chicago economist and CDE co-founder Glen Weyl, who has proven with University of Chicago statistician Steven Lalley that it is the only optimal and practical way to make collective decisions.

[Warren D Smith]

Major problems with "Quadratic Voting" for most practical purposes are
discussed here:
http://www.rangevoting.org/MonetizedRV.html

The webpage Clay cited says to use "credits" rather than real money to
buy votes,
a suggestion I had made (perhaps others had also made it) which perhaps
is an improvement. The collectivedecisionengines.com
place seems really weird, whatever they are. Maybe they'll
accomplish something?


--
Warren D. Smith
http://RangeVoting.org <-- add your endorsement (by clicking
"endorse" as 1st step)

[Warren D Smith]

http://collectivedecisionengines.com/quadratic-voting.html
INTERESTING QUOTE
All existing systems, such as one-person-one-vote, approval voting, or
surveys, leave opportunities for _everyone_ to be made happier without
_anyone_ doing worse.
END QUOTE

What did that mean?
I suppose you could give money to people to make them happier, but
that would require taking money from other people, making them less happy
i.e. "doing worse." You could change the election winner, which again
would make
some happier some unhappier.

So I guess what they must have meant is: if & when some system elects
a regrettable winner then (a) change the winner to one with better
utility-sum, (b) whomever that
change made unhappy is monetarily compensated until they become happier
in net, (c) whomever the winner-change made happier, has to pay some money,
but not enough that they in net become unhappier -- and since the
winner-change increased summed-happiness, if "happiness" and "money"
are linearly related --
and with constant proportionality factor independent of who --
then it always is possible to make EVERYBODY happier in this way.

Nice observation.

BUT now suppose the proportionality constants depend on the person,
e.g. it takes $1 to give you the same happiness increment as I get
from $2 (e.g. because I was richer
to start with). In that case, is there still a way to make everybody
happier? NO.

Counterexample: suppose there are 2 voters (you and me) and 2 candidates
(for simplicity, say also: you & me). We both want to win. Overall
summed happiness maximized if you win. But I win. So we change the
winner to you. Making you $10
happier and me $19 unhappier. But because our happiness/money
proportionality constants differ in 1:2 ratio we are overall happier.
But no matter how much
you now pay me, no way we both end up happier than if I had won.

Another Counterexample with perhaps a little more irritating impact is
the "kill the Jews" vote, where the question is, should we kill all
the Jews (a minority) take their money, and redistribute it to
everybody else? Honest simple majority vote unfortunately kills the
Jews.
("Honest" meaning all voters taking only their own interests into
consideration when voting.
You know, the way Economists say we act.)
Taking all the money from the Jews and redistributing it among the nonJews, but
not killing them, does not seem to make everybody happier (?),
contradicting the Claim... although it at least seems not to make
anybody unhappier.

AND BUT now suppose nonlinear relation between happiness and money
(but same relation for everybody). Again one can easily find numerous
counterexamples...

[Warren D Smith]

>So I guess what they must have meant is: if & when some system elects
a regrettable winner then (a) change the winner to one with better
utility-sum, (b) whomever that
change made unhappy is monetarily compensated until they become happier
in net, (c) whomever the winner-change made happier, has to pay some money,
but not enough that they in net become unhappier -- and since the
winner-change increased summed-happiness, if "happiness" and "money"
are linearly related --
and with constant proportionality factor independent of who --
then it always is possible to make EVERYBODY happier in this way.

>Nice observation.

--No, wait. Even under these assumptions, the "nice observation" still remains
untrue unless we allow people to go into arbitrary debt.

[Warren D Smith]

>>So I guess what they must have meant is: if & when some system elects
> a regrettable winner then (a) change the winner to one with better
> utility-sum, (b) whomever that
> change made unhappy is monetarily compensated until they become happier
> in net, (c) whomever the winner-change made happier, has to pay some money,
> but not enough that they in net become unhappier -- and since the
> winner-change increased summed-happiness, if "happiness" and "money"
> are linearly related --
> and with constant proportionality factor independent of who --
> then it always is possible to make EVERYBODY happier in this way.
>
>>Nice observation.
>
> --No, wait. Even under these assumptions, the "nice observation" still
> remains
> untrue unless we allow people to go into arbitrary debt.

--I think I now see what they must have meant:
CLAIM:
If money and happiness linearly related, same relation for every
person, then there
are elections (maybe few, but a nonempty set)
where there is a way for everybody to be made happier.
END CLAIM.

In other words: you name a voting system (e.g. approval voting) and
then I can construct some artificial example election with some
artificial voter and candidate sets in which it would be possible to
make everybody happier by changing the winner and redistributing
money.

Indeed, it seems to me this is true even if the money-happiness
relation is any monotonic increasing smooth curve -- because I will
make my artificial example involve all people
with same wealth, and only very tiny money perturbations so the smooth curve
becomes effectively linear.

[Clay Shentrup]

I view these folks the way James Randi views Uri Geller.

[Warren D Smith]

"Quadratic voting" and its ilk may have some use, so I would not
quite call it "Uri Geller" -- but it has been greatly overhyped, and lied
about too. And the very idea that now this company is going to SELL
a pretty trivial voting method somehow (apparently their
game plan is to keep it very mysterious and impressive sounding,
kind of like Long Term Capital Management... nothing you can do yourselves,
you need to pay experts like us) seems pretty wacky
to my poor naive non-economist mind.

On the other hand, if they prove me wrong, that'll be nice.