Brought out for special thought:
rating is easier than ranking. You can express this
computationally, by saying that ranking requires O(n²) pairwise
comparisons of candidates (or perhaps for some autistic savants who
heap-sort in their head, O[n log(n)]), while rating requires O(n)
comparisons of candidates against an absolute scale. You can
express it empirically; this has been confirmed by ballot spoilage
rates, speed, and self-report in study after study.
This somehow does not fit as to rating vs ranking. I look at A and B,
doing comparisons as needed, and assign each a value to use:
. For ranking the values can show which exist: A<B, A=B, or A>B,
and can be used as is unless they need to be converted to whatever
format may be acceptable.
. For rating the values need to be scaled.
Thus what needs doing is a trivial bit of extra effort for rating.
The comparison effort was shared.
"Ballot spoilage rates" also puzzle. Where can I find what magic lets
non-Condorcet have less such than Condorcet, for I do not believe such
magic exists, unless Condorcet is given undeserved problems.
On Aug 5, 2011, at 8:57 AM, Jameson Quinn wrote:
If I understand Robert correctly, I think that his concern is to
find whichever system gets good results, while still leaving the
voters' job crystal-clear. In particular, the first priority would
be to find a system which puts the minimum burden of strategic
thinking on the voter. Insofar as it's consistent with this first
priority, his other priorities are to simplify the ballot and to
elect a better winner.
He says that he finds Condorcet to beat Range, Approval, IRV, and
Plurality on those priorities. I can't say I disagree. There is no
question that the average voter can get away with less strategizing
under Condorcet than under any of those systems. Range and Approval
both give significantly more power to a voter who can correctly
guess which two candidates are the frontrunners; and a conscientious
voter simply can't leave it up to the law of averages to balance out
that power between themself and those crazies on the other side. And
IRV and Plurality both require lesser-evil thinking and favorite
betrayal, which is worse.
But I would contend that, just as clearly, both Majority Judgment
and SODA beat Condorcet by these criteria. These are two very-
different systems, so I'll tackle them separately.
Why is Majority Judgment easier on the voter than Condorcet? First
and foremost, because rating is easier than ranking. You can express
this computationally, by saying that ranking requires O(n²) pairwise
comparisons of candidates (or perhaps for some autistic savants who
heap-sort in their head, O[n log(n)]), while rating requires O(n)
comparisons of candidates against an absolute scale. You can express
it empirically; this has been confirmed by ballot spoilage rates,
speed, and self-report in study after study. You can look at the
trend of websites - Yelp, IMDB, Amazon, and an interminable etc.,
all use ratings. HotOrNot explicitly experimented with FaceMash-
style comparisons and found that even a single two-way comparison
was harder for users than ratings.
Second, MJ has, if anything, less of a strategic incentive than
Condorcet. Balinski and Laraki found that, in a simulation seeded by
real polled ballots from the 2007 French presidential election, the
MJ result was more likely to be stable under strategy than the
Condorcet result. Warren Smith has argued that, if you know who the
two frontrunners are, a Condorcet burial strategy is essentially
risk-free, while saving cognitive effort. In MJ, on the other hand,
chances are that your honest vote is already getting all possible
benefits from such a strategy.
The last two paragraphs may seem counterintuitive to people who are
used to thinking of ratings ballots in Range terms. That's because,
in Range, some amount of strategic thinking is almost totally
inevitable. It would be ridiculous to mark a Range ballot by simply
rating each candidate on an absolute scale, without normalizing so
that you marked at least one max and one min vote. And voting power
continues to increase as you move towards an approval-style ballot.
With MJ, on the other hand, even a non-normalized vote with no max
or min ratings could well have full strategic voting power.
As to ballot simplicity, it's partly a matter of taste. I'd say that
MJ has the edge over Condorcet, but you might disagree. Of course,
it would be possible to do Condorcet with a rated ballot, but since
I don't know anyone who seriously advocates that, I'm neglecting
that possibility.
Condorcet can use THE SAME values as rating - the ratings show what is
needed to identify A<B, A=B, and A>B.
For results, I think that MJ and Condorcet both do more than well
enough. Is it better to have the candidate who makes some majority
happiest, as in MJ, or the one who has majorities over all the
others, as in Condorcet? On principle, it's hard to say. In the end,
I definitely sympathize with Warren that BR is the best measure of
results (while disagreeing that results should be the end-all
criterion). I haven't seen BR results for MJ, but based on its
similarity to Range and MCA, I'd suspect MJ clearly beats Condorcet
by this measure.
If you want to read more about MJ, the place to start are the papers
on Laraki's home page. These are pro-MJ papers, but honestly, I've
found that I got more understanding even of MJ's flaws from reading
these critically, than I did from reading the couple of shoddy anti-
MJ papers that you'll find in a Google scholar search.
....
As to SODA, the case is also clear. Less need for strategic thinking
than Condorcet? Check. Simpler ballots? In spades. Good results? If
voting blocs tend to be well-defined so that voters agree with their
favorite candidates — not too excessive an assumption in an
ideological, partisan election — then the CW will win in SODA more
often than in Condorcet! That is to say, honesty will be the
strategic equilibrium in SODA, and lead to a CW win; but in
Condorcet, strategy could confound this. In practice, I think that
the only case where the CW will not win SODA is when they're an
relatively-unknown centrist. Such a person can become the CW
precisely because voters do not know their flaws; but in SODA, the
other candidates will evaluate that person much more carefully
before they transfer a winning pile of votes to an unknown.
....
I've been actively interested in voting systems for going on two
decades now, and sometimes it's frustrating. We continue to argue
amongst ourselves between system A and system B, we continue to
invent new system Z, while the rest of the world basically ignores us.
In that time, it seems as if I've favored everything under the sun.
If I remember correctly, I've gone from Borda, to IRV, to Condorcet,
to Range, to Approval, to Asset, to Bucklin, to Bucklin-which-I-
called-MCA, to MJ, back to Approval, to SODA. So I think I can
pretty fairly say that I'm not fixated on any one system. I also
think that I understand the arguments in favor of different systems.
And I honestly have more hope on this issue right now than I have
for some time. SODA could be the winning combo. Yes, perhaps I'm
biased, because I helped invent it. But if you look at that list of
what I've supported, you won't see any other systems where I'd done
more than minor tweaks around the edges. And that's not for lack of
opportunity; I've invented many systems of my own. It's because I
always wanted to back a system which had other supporters, because
we're not going to accomplish anything each working alone.
What's so great about SODA? I think it would entirely liberate
voters from having to think about strategy. I know that it is the
simplest possible task for the voter. And I expect that it would
give great results. These are R B-J's three criteria, as I
understand them, and I agree with them. I know that other people
here might put them in a different order; for instance, the Range
advocates would probably put results (that is, BR) first.
SODA has problems, too. It's an obscure system invented yesterday.
And it doesn't even really lend itself to publishable results; it's
hard to prove anything about it or even do a monte-carlo simulation
without making questionable assumptions about how much or how little
voting blocs agree with their favored candidate. Finally, it cannot
work unless the candidates are human beings; it can't evaluate
options without attaching them to spokespeople. But those problems
are can be addressed, one way or another, with enough time and effort.
I'd still enthusiastically vote for almost any reform to Plurality -
certainly any of the systems except Borda that I listed as having
supported. But I have hope that EM researchers and advocates can
start coming together as we never have before. I'm not saying that
everyone will agree with me. I wouldn't even want for debate to
stop. But I think that now is the time to start laying the first
foundations of a new unity, and if you ask me right now, SODA has
the strength to be the cornerstone.
I hope this does not sound too arrogant. Certainly, I am humbled by
the many tasks before us. And no matter how rousingly optimistic my
rhetoric, experience reminds me that getting even two of us to agree
on anything is always harder than you think.
But still; I remain:
Hopefully y'rs,
Jameson Quinn
2011/8/4 robert bristow-johnson <[email protected]>
On Aug 4, 2011, at 3:20 AM, bob wrote:
--- In [email protected], "thenewthirdparty"
<thenewthirdparty@...> wrote:
Guys and Gals,
I now see Range Voting as a very important component to getting
third parties elected. But I don't see how the Range Voting group
will ever change the minds of the public in order for it to be a
reality.
and they haven't changed my mind about it, even though i'm not
opposed to election policy reform nor of moving past FPP. i fully
recognize why the simple vote-for-one ballot (either FPP or delayed-
top-two-runoff) disadvantages third-party and independent candidates.
this was a point i brought up during in Burlington IRV debate: one
of the vocal opponents to IRV was, 3 years previously, a minor
candidate for mayor in Burlington Vermont. i would almost say a non-
serious candidate, but he got on the ballot (his name is Loyal
Ploof). now he lost to the Prog candidate who was elected in 2006
and he was a sorta anti-establishment rabble rouser (if he could get
a rabble).
now (i told them this), suppose i'm standing in the rabble and Loyal
says something that we all sorta know but the contending candidates
aren't gonna bring up and i hear it and i say "yeah, right! Loyal's
right!" maybe even he's a largely single-issue candidate, maybe
not. but i want to send a message to city hall by voting for Loyal
but the election between the real contenders might be close and my
two-party contingency candidate may need my vote. so Loyal doesn't
get it, because even if i agree with him and *want* to vote for him,
i dare not.
it's the typical Spoiler problem, that discourages voting for third-
party or independent candidates. if they can never sufficient vote
(because the race between credible candidates may be close) third
parties cannot get off the ground and become contenders. but i was
surprized that this guy who would directly benefit from a ranked
ballot would be opposed to it. (he didn't like the Prog mayor and
essentially jumped in the boat with the other Prog-haters that
believed, falsely, that IRV specifically favored the Progs in
Burlington.)
that said, and to repeat that i also understand IRV to have *failed*
in Burlington in 2009, i am not at all impressed with Range or Score
voting for governmental elections (for certain Olympic sports, sure,
but not for governmental elections). one of the complaints we have
against both FPP and IRV (as we found out in Burlington in 2009) is
placing obvious burdens of tactical voting on the electorate. we
don't *like* having to forsake our favorite candidate in order to
accomplish some other political imperative. FPP discourages the
Nader voters from voting for their favorite candidate in 2000 by
punishing them when it became clear that their vote cause Bush to be
elected. and IRV discourages the GOP Prog-haters in Burlington from
voting for their favorite candidate in 2009 when they discover that
marking their favorite as #1 on the ballot actually caused the Prog
to win.
now, it's not the ranked ballot that failed these voters, it was the
Hare-STV method of tabulating the vote. Condorcet would have taken
the same ballot data and elected the candidate that was preferred by
the electorate over any other specific candidate. The GOP who lost
the most in the election would neither have gotten punished for
their sincere 1st-choice vote (if IRV had survived, in 2012 these
guys would be saying to themselves in the polls: "I gotta choose
between Liberal and More-Liberal, because if I vote for the guy I
really like, More-Liberal gets elected"), they would have been more
satisfied with the Condorcet winner than with the IRV winner, who
was their least favorite. And the Progs would have been more
satisfied with the Condorcet winner than with the apparent FPP
winner (the GOP), but they would be unhappy with the result due to
rivalry the Progs and Dems have for the common liberal voter in this
town.
Ranked-choice voting requires less strategizing by the voter than
Range because it requires less information. with a ranked ballot,
all the voter needs to decide is who, in every contingency that
matters to the voter, who he or she would vote for. they don't need
to decide how much *more* they like Mother Teresa over Ghandi. If
they really want to bury a third candidate, Stalin, they have to
sacrifice their preference between the two virtuous and the election
might be decided between them. Or maybe the election will turn out
to be a battle between Stalin and Satan and they might rather live
under Stalin than Satan, so they want to bump him up a little (leave
Satan with a score of 0). but what if Satan wins because not enough
voters scored Stalin up enough? or what if either Teresa or Ghandi
lose to Stalin because too many voters scored Stalin too high (for
fear of electing Satan)?
what to do? what to do?
but a ranked ballot is easy:
Teresa > Ghandi > Stalin > Satan
or, if you're more Hindu than Christian:
Ghandi > Teresa > Stalin > Satan
no tactical thinking necessary for the ranked ballot when it decided
by Condorcet and a Condorcet winner exists. and, if a CW exists,
the result is perfectly consistent, in every contingency, with the
simple-majority, two candidate, one-person-one-vote election that
everyone is familiar with.
Does someone have thoughts on how to get your Range Voting plan
voted into action? I would like to hear how Range Voting moves
beyond more than just a good idea.
how does it move beyond "good idea" when it hasn't advanced to that
square? (sorry Warren, i *really* have a lot of respect for you and
your scholarship and your Burlington IRV page at your website, but
you're still not convincing regarding Range. a little more
convincing regarding Approval, but i would still not support that
for political office, maybe the judiciary or some boards, but not
executive nor legislative.)
listen, everybody agrees with how a simple 2-candidate election
should be decided: person with the most votes wins and every voters
vote is of equal value. "simple majority" and "one-person-one-vote".
wouldn't it make a lot more sense, since IRV is discredited, and FPP
is clearly flawed, to put your energy into educating people about
what goes wrong and *has* gone wrong in those elections and present
an alternative with ballot no more complicated than with IRV and
truer to the hypothetical 2-person race, whether the spoiler runs or
not?
I think we need to start a PAC or even maybe a party that has the
sole objective of getting rid of plurality voting.
doesn't one exist? why not team up with FairVote?
We need to be able to communicate that competitive elections in
which there is no vote splitting is the most important thing we can
do to hold politicians accountable.
sure, and how does Condorcet cause vote splitting? you don't need
Range to address the problem of splitting the majority vote.
We also need to be willing to vote for candidates who support
getting rid of plurality regardless of what other positions that
candidate holds.
oooh, i dunno if i can handle that. weirder things have happened
than that of Michelle Bachmann supporting ranked-choice voting. i
wouldn't vote for her even if she *loved* Condorcet.
We need to communicate that once we get over this hump, we will no
longer have to worry about having to vote for the lesser of two
evils ever again.
Another thing we can do is email and tweet news hosts like Rachael
Maddow and ask them to do a segment on different voting systems. If
we organize to tweet pundits at the same time, maybe they'll get the
message.
dunno who Rachel Maddow is. guess i better google her. how about
Chris Matthews?
On 8/4/11 9:16 PM, Dave Ketchum wrote:
Here I talk of moving up from FPP to Range or Condorcet. I do not
get into other single-winner elections or into multi-winner
elections - while such deserve considering, they distract from my
primary goal, which is to promote moving upward without getting
buried in details.
Voters should see advantages in moving up to a better method.
To vote for one, as in FPP:
. In Range, assign your choice a maximum rating.
. In Condorcet, simply rank your choice.
which is simpler?
Voting for two is using more power than FPP offers. Often there is
a major pair of candidates for which you prefer one, and one other
that you also want to vote for: For your second choice you could
give the same rank or rating, or lower:
. In Range you assign first choice maximum rating. Unrated
share minimum. The farther you rate second below max, the stronger
your vote for max over second. BUT, the nearer you rate second to
unrated, the weaker you rate second over unrated.
. In Condorcet, rank your first choice higher than your second.
ditto.
Voting for more is doable:
. In Range your difference in rating between any two is how much
you prefer the higher over the lower, and the sum of these
differences decides which wins their race.
. In Condorcet they count how many rank A>B vs how many rank B>A.
which meaning complies more with equal weighting of each voter's
vote (what we normally mean by "one-person-one-vote")?
Politicians may hesitate in moving up to more powerful methods.
Range or Condorcet can cost more, but getting a truer reading as to
voter choices can be worth the pain.
i'm sorry, guys. i'm really sorry, Warren, but between Condorcet
and Range, it just ain't close.
--
r b-j [email protected]
"Imagination is more important than knowledge."
----
Election-Methods mailing list - see http://electorama.com/em for list info
