[Election-Methods] Election-Methods Digest, Vol 42, Issue 80

[CLAY SHENTRUP]

> Date: Mon, 31 Dec 2007 15:30:31 +0200
> From: Juho <juho4880 at yahoo.co.uk>

> Greater social utility means greater average social utility per voter
> but that does not mean greater average _personal_ utility nor that
> this voter would achieve better personal utility.

a voter's expected _personal_ utility is the social utility divided by
the number of voters.  see the part about harsanyi here:
http://rangevoting.org/UtilFoundns.html

> (That voter's personal utility might be in conflict with the overall
> social utility.

in a particular election, of course. and in a particular election, the
world's worst voting method might give you a better result than the
world's best voting method.  but in any given random election, the
higher the social utility, the higher your expected personal utility.
that is, while you may have specific knowledge about a particular
election that tells you you'd be better off with the worst system,
over the course of many elections, you'd rather have the method with
the greatest _social_ utility.

your flaw here is the same as when a gambler makes the strategically
optimal move, but then loses and says "i should have used the other
move."  well, no you shouldn't have, because the next time you're in
the same situation, you'll better off maximizing your expected value -
even if the strategy which has that effect screws you over in a
particular circumstance.

there may be elections where you'd have been happier with plurality
than with range voting - but they will be rare.  more often than not,
the opposite will be the case.  and you have to think about expected
value.

> (I wonder if you would vote Range=90, Condorcet=30, Plurality=10 or
> Range=100, Condorcet=20, Plurality=0 or Range=100, Condorcet=0,
> Plurality=0.)

it's hard to say, because it depends on how many candidates typically
run.  if less than 4, than surely condorcet is better than plurality.
but with 4, if there is enough strategic voting, you can get the DH3
pathology, and have a worse result than with plurality.

i just compare bayesian regret figures under a scenario most closely
resembling the election in question.

> Having the approach "all others please vote sincerely so that I can
> then vote strategically and have more weight" doesn't make Range a
> very nice method.

you're confusing two different concepts.  a voter wants to vote
strategically and have other voters vote honestly under just about
_any_ method - it has nothing to do with range voting.  in fact, the
relative benefit a voter gets by having everyone else vote sincerely
(against their own best interest) is less with range voting than with
other methods.

the point is that with range voting, that voter will have a better
expected happiness than with other methods - especially if the other
voters are strategic!  which is precisely the opposite point to what
it seems that you perceived me to imply.

> Yes, in many situations Range would indeed provide good results. The
> problem is that in competitive situations it may not do so.

no!  warren's utility figures tested the full range of 0% strategic
voters to 100% strategic voters, and found that range voting was
superior to the other common methods.  in fact, range voting's
superiority over rivals such as condorcet and borda actually
_increases_ the more "competitive" the election becomes.
http://rangevoting.org/StratHonMix.html

how is this not more well-known in a discussion list of this sort?

> (Full degradation to pure Approval gives probably better results than
> partial degradation.)

no it doesn't.  the more voters vote honestly rather than
strategically, the better the net utility.

> Different elections have also different needs
> => maximizing the average utility is not always what we want.

name an example.  that sounds preposterous.  utility is _by
definition_ the measure of what voters want.

> Do you say that in Condorcet elections the best strategy for A
> supporters would be to vote A>X>Y>Z>B if they believe that A is the
> strongest candidate and B is the second strongest candidate?

yes!  it may be a little more nuanced than that, but that's the gist.
warren can explain it better i'm sure.

and whether it's a good idea or not, a lot of people will just
intuitively do it, because it "feels" like they are pushing B down
more strongly.  and they are sort of right.

> I just picked two strategies that have been often recommended and
> that are simple.

what you described for condorcet was not a "strategy" -- it was a sincerity! :D

> Many Condorcet strategies are not easy to
> understand and apply ((many of them are not very efficient and safe
> either)) and are therefore not good candidates for this comparison.

that's nonsense.  you've been lied to.  raise your favorite
front-runner to the top and bury the other front-runners.  sound
complicated to you?

class dismissed.

CLAY