[Election-Methods] more rob brown confusion

[CLAY SHENTRUP]

>> it sounds like you're saying they're not relevant because they're
>> rare. but in a recent look warren took at around 150 australian irv
>> elections, only 8 would have been differently with plurality (and
>> assuming more voters would have strategically not voted for their
>> sincere favorite if using plurality, plausibly more like 4 instead of
>> 8).
> Based on what? Can I assume he used his patented DNA computer to model the
brains of the voters?

based on the actual election results.  only 8 elections would have been
different with plurality, with people voting for the candidate they
top-ranked with IRV.  and plausibly with plurality, more of them would have
strategically voted for their favorite major party candidate, reducing that
8.

> Ummm, for someone who likes to accuse people of not attending election
101,
you seem to be using the most naive argument yet. The tired "your vote is
unlikely to make a difference so you must be voting out of altruism so
therefore you will not try to manipulate it in your favor" argument.

i'm arguing the _opposite_ of that.  people _will_ manipulate it in their
favor, even if it is unlikely to make a difference.

> For what it is worth, that is one reason I kept trying to get you to
discuss
the one kind of election where this *isn't* an issue, which is a "vote for a
numerical value" situation (using interpolated median)

that's not the only kind of election where strategic manipulation isn't an
issue.  we could also use RandRank - let voters rank the candidates, then
randomly pick 2 of them for a head-to-head matchup.  and at least then we're
talking about something related to the election of discrete entities.

> But you refuse to listen to such things because you claim you are only
interested in "real world" elections. (and then I am expected to listen
when you make a contrived hypothetical case with 3 candidates and only 3
voters)

but those cases aren't _fundamentally_ different than large elections.  they
are a smaller version of the same thing.  i could multiply them by millions
to make them more realistic, but then it would be vastly harder to focus on
the fundamental axioms we're debating.  again, my problem with your "voting
on a number" thing is that it's fundamentally different, because it's
totally one-dimensional, with infinite options, and with those options
extending to infinite extremity in both directions.

and more to the point, median voting is almost definitely not the best
system.  with all of the other possible ways you could tweak it, there's
very probably some even better method, like "average of the median of the
bottom half and the median of the top half".

> Regardless, you can't assume that those who vote stregically under
plurality
are the same as those who would vote strategically under range or condorcet
or any other method.

in very broad general terms, no.  in fact i believe that range voting will
inspire greater honesty, so i agree with you there.

however, if we look at specific examples like the 2000 election, it's clear
that anyone who had strategically voted for gore, but preferred nader, would
still have an incentive to give gore a "10" under range voting, especially
since he wouldn't have to throw nader under the bus to do so.  it would be
very strange to find an individual who voted strategically under plurality
(was willing to throw nader under the bus to maximize gore's chances of
beating bush) but would not be willing to do so if it _didn't_ require him
to betray nader.

> Sorry, did I give the impression I learned something from you?

i've corrected you on many things, but whether you've learned anything is
questionable.  generally it's in one ear and out the other, and i can point
to numerous examples that prove this, where you've demonstrated ignorance on
something that i already explained to you.

> For what its worth, my article on movie voting which has been on the web
for
2 years states pretty clearly that the DSV system described in it is
identical in effect as condorcet.

and condorcet is gameable, so...

> Regardless, I think a DSV system affords the most possibility of making it
100% strategy proof, because by definition the software agent is using
optimum strategy on your behalf. You still haven't explained how you can
improve on optimum strategy by lying to the agent, unless the agent really
isn't 100% optimal.

if there's a condorcet cycle, the agent will NOT act on the behalf of the
majority who prefers a candidate other than the one who wins the cycle
resolution.  then many will wish they had lied, to get their second rather
than their third choice among the smith set.

> BTW, you told me that I was the first one who had made a convincing case
for
"fairness" being anything other than the sum of net short-term utilities,
and being something that you might actually want to factor into future
calculations. That seems pretty elementary. Or maybe pre-schoolery.

you keep saying short-term or long-term, which is besides the point.  the
point is that there are _externalities_ aside from the candidate
satisfaction, like how much the election costs, or what the ballot spoilage
rates are, or whatever.  some of these are actually short-term, just as much
of the utility of the election outcome itself is long-term.  i mean, i'm
still suffering from fdr's new deal socialism years after his election.  so
i'd like to see you keep the focus on "utility from the election result" vs.
"utility from secondary effects related to the voting method".

in any case, i acknowledged that you had a point that there are
externalities.  i've reaffirmed this many times.  but i still haven't seen
you demonstrate any plausible externalities which would overcome those
election result utilities.  you say that many people would find range voting
unfair.  but i don't see nader supportings saying it's unfair for other
nader supporters to strategically vote for gore.  if they want to be
strategic as well, they have the option.

> First, I would argue that doing so will not advance their interests if
looked at in a statistical sense, especially if they know that other voters
are likely to be doing the same thing. It becomes too hard to predict what
others will do.
> Even so, if that *is* possible then the agent isn't operating on the
voter's
behalf, and its algorithm should be changed to have the same effect when
casting the approval ballot.

it doesn't matter what algorithm you use.  in a condorcet cycle, no matter
which candidate the algorithm picks, there will be a majority who prefers
another candidate, and if it is really to operate on their behalf, it must
cycle forever, never picking a winner.

welcome to arrow's theorem.

> And I never said sell it for
political elections, not in the short term. I'm more interested in getting
people to use it for non-government elections for now (organizations, etc),
so that it can get some real world testing.

i agree that methods like range and condorcet should be tested much more, to
see more, but it's disconcerting that you can't actually read voters'
minds.  so you can't tell how many voters were strategic, and thus you can't
tell how well the voting method actually worked.  which kind of renders all
that testing moot.

> And its not like Range voting has a snowball's chance in hell of being
used
for political elections, not in our lifetimes.

i adamantly disagree.  approval voting, the simplest form, has already seen
vastly more use than condorcet.  it was used in venice for hundreds of
years, it's used at san francisco state university, it's used to elect the
secretary general of the u.n.  there were recent articles on range and
approval in newsweek.  there's a book coming out in favor of range voting,
by a bestselling author, which has been positively reviewed by kenneth arrow
himself.  there are all kinds of reasons to believe it's plausible to
implement it in our lifetimes.

> Sorry to say, but just
telling people that you've proven that they want range, even though they
don't think they do, isn't gonna do it.

you're right.  you have to show how it helps them get better election
results, and reduces ballot spoilage and things like that.

> Or not. Range and approval simply converge to Condorcet as people are
better informed anyway, you said it yourself.

but to the extent that they are strategic enough to make that happen,
condorcet converges into something worse.  and if they are honest enough to
make condorcet really _be_ condorcet, then they are probably honest enough
to make range voting even better.

> The only difference is that
Range and Approval have shifted the responsibility for resolving cycles to
polls, the media, and psychology, rather than having as much as possible of
cycle resolution in the tabulation method.

that doesn't make any sense.  maybe what you meant is that range requires
voters to try to figure out which vote is in their best interest, whereas
condorcet does it automatically.  but then, that's not true, because
condorcet is also susceptible to strategy.  so this comment still makes no
sense.

>> range voting is better and much simpler/cheaper to adopt than
>> condorcet, by far. (better as in, picks more representative
>> candidates.)
>Not better as in "more fair" though.

it's much more fair.  a voter's expected happiness with the election is
greater.

you suffer from the strange belief that it is more fair to reduce a voter's
expected happiness in order to guarantee him that his vote will have as much
effect when he cares very much as when he cares less.

that is economically irrational.

> Tell me where in Princeton math PhD Warren's simulations does he
demonstrate
that the long term value of perceived fairness doesn't outweigh whatever
short term value having more "representative candidates" has?

you say "perceived" fairness.  does that mean you acknowledge that range
voting is actually more fair even though condorcet might be _perceived_ to
be more fair by more people?

> All I've heard you say is that you doubt that fairness is important enough
to people to outweigh whatever gain in short term utility you think they
will get from range voting. (meanwhile you sure seem to love that book,
subtitled "why elections aren't fair")

and that book says, "A system called range voting, already widely used on
the Internet, is the fairest voting method of all.."

i agree.  i don't think condorcet is more fair to me by reducing my expected
satisfaction with the election result.  i'll sacrifice some "power" in
contests that mean less to me in exchange for more power in elections that
mean more.  that is what any rational voter would want to do.

so are there enough irrational voters to riot in the streets if they don't
realize that?  i doubt it.  and you sure haven't proven it.

>> it can be done on ordinary voting machines and reduces
>> spoilage rates. if you think advocating range voting is preserving
>> the status quo, then jousting at the condorcet windmill should make
>> you feel like the ultimate advocate of the status quo.
> I'm not sure how you see that I am jousting at the condorcet windmill. I
like condorcet, so why would you say that?

well, maybe you're being less quixotic if your goal isn't so much to get it
used in government elections.  but why would you say "i like condorcet."?

why would you choose to have less satisfying election results just for the
faux privilege of always having equal discriminating power in all
head-to-head matchups?  that's a losing gamble.

> And for what its worth, I support IRV (God forbid), in the sense that it
has
momentum and does address some of the worst problems with plurality.

the "has momentum" argument is also a great reason why we should support
global warming instead of global cooling.  all that extra momentum means
global warming is the way to go!

see, here's a case where externalities come into play. irv is more
utilitarian than plurality.  but it also causes 7 times as many spoiled
ballots, and catalyzes the adoption of fraud-prone electronic voting
machines, and cannot be counted in precincts.  so it actually may be worse
than plurality.  it's hard to be sure, but i'm just saying it's not a clear
improvement.

> Yes I am well aware it is far from perfect, but I would rather it gain
momentum,
and hopefully once people are more comfortable with the whole idea of
ranking candidates they will be more receptive to other methods of
tabulation. So my main effort is to advocate "ranked voting methods",
condorcet being my favorite, IRV being the one likely to make short term
gains. I know your approach is to bullet vote your favorite and screw the
consequences, so this strategic approach is probably a bit tough to grasp
for you.

if i supported condorcet (i.e. wanted to shoot myself in the foot), then i
could "grasp" it.  but since i think it's imperative to get away from ranked
voting methods, for my own personal happiness (and yours too), and since irv
may actually be worse than plurality, i cannot "grasp" this.  we need to be
focusing on approval voting in the immediate.

> I think the Power Rangers are obnoxious and rude and misguided and do
nothing to advance voting reform.

the reality is we're enlightened and scientific and are doing more than
anyone to educate people about the dangers of their misguided pro-rank-order
beliefs.

that includes a very misguided rob brown, who thinks people find it more
fair to be less happy.

> And I think their method is stupid, because it is just approval voting
with
some attempted deception thrown in for good measure.

that deception is good.  it gets lots of voters to reveal more information
about their preferences than they would with approval voting.  hence a more
satisfied average voter.

> And because it is based on the idea that short term utility is the only
criteria for measuring something.

no it's not.  i think range voting has numerous long-term benefits too.

> And because it is clear in its very design that it is all about rewarding
those who discourtiously exaggerate

1. so does _every_ deterministic method.
2. even if many more people are strategic with range than other methods, it
still achieves better average voter satisfaction with the election results.

this has been explained to you before, so stop being a broken record.

> (yes I am aware the condorcet can, in some cases, reward those who vote
insincerely, but I think there is a difference between leaving the gate wide
open (Range) vs. a system that makes you climb over a tall fence with razor
wire (condorcet).

there's no evidence that analogy is accurate.  it actually appears that
condorcet gives a bigger incentive to exaggerate.  and even if that wide
open gate lets a lot more intruders in, range renders them more impotent to
cause damage.
http://rangevoting.org/StratHonMix.html

>> you want a radically more complex
>Complex if you only count the clock cycles of the computers. Not if you
count the confusion that will go on in the brains of the voters.

if you really believe that condorcet is harder to game, than it causes even
MORE confusion in the minds of the voters.  duh.

> Your opinion. I know you think it is fact. It's not. Unless "short term
happiness with the actual results" is equal to "long term satisfaction with
the results, the method used, and whether it seems fair and equitable".
But, of course, it's not.

range is more fair than condorcet.  there's a greater average happiness with
election outcomes.

your argument is, no it's not really more fair when you consider how unhappy
many people will be that it was less fair.  think about that for a moment.
resist the urge to reflexively respond on auto-pilot, and actually think.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20080102/db3b4111/attachment-0002.htm>