[EM] (with paragraphs this time) Comparison of IRV, AIRV an Approval for small groups needing an easy handcount.

[Michael Ossipoff]

Sorry--It posted without paragraph spacings. I'm going to try to use
periods (".") to hold the blank spaces: Also, I've noticed that it
wasn't set for plain text, and that setting might avoid the problem.
.
.
Meeting MMC, but having chicken dilemma, Bucklin and Beatpath would
serve ok for mutual majorities who were completely mutually trusting
and trustworthy.
.
As for Beatpath, there's little point for it, because there are
equally easily-counted, and more easiy-counted, methods that meet MMC
and _don't_ have chicken dilemma.
.
.
Bucklin has a better chance of justification, because its handcount is
easier than that of the pairwise-count methods.
.
I agree with those here who argue that computer-counting can be fully
secure, and we've been discussing that. But some organizations and
temporary small groups who have a collective choice to make might not
have a computer, with coun-software, conveniently available. Of course
they could conduct their voting at the CIVS website, but suppose that,
for some reason that isn't feasible either.
.
..
I'm talking about such a small organization or temporary group. I was
agreenig that electing the CW is good for stability, and avoidance of
a displeased majority. But suppose that small group must do
handcounting. Pairwise-count methods are computation-intensive. The
idea is to look at all of the pairwise preferences that you express,
so as to take them all into account, let you vote all of them, for the
purpose of finding the best compromise that you can get, to avoid a
dis-satisfied majority. But, by the nature of its goal, that's
computation-intensive.
.
.
So, replacing IRV with Condorcet-IRV, Woodall, or Schwartz Woodall
might not be feasible for the small group that I'm talking about.
Automatically looking at every pairwise preference is out. That leaves
two possibilities that I suggest:
..
.
1. Just do IRV. You loose the Condorcet efficiency, and IRV can choose
from a MM-preferred set in a way that violates the wishes of some
majority, causing majority dis-satisfaction. But that isn't so bad, if
it's for a good reason, such as the unfeasibility of a pairwise count.
If it's a political poll, such as the current one at CIVS, maybe you'd
prfefer CC compliance, but it isn't essential. The Greens offer IRV in
their platform, and I have no objection to not recognizing every
majority, but only the mutual majorities, as does IRV. And if it's ok
for actual (Green scenario) official public elections, then it's ok
for political polls too--unless you have the specialized purpose of
looking for the best that your faction can get, by finding the CW.
.
I've said that, for amicable organizations, IRV is probably too
uncompromising, too adversarial. Sure, but if Condorcet-IRV, Woodall,
and Schwartz Woodall aren't feasible, because pairwise counting is too
computation-intensive, the that's reason enough to say "That's ok."
After all, you or your group won't blame yourself or theirselves,
because giving up CC wasn't done by choice.
..
2. If it isn't feasible to count every one of everyone's pairwise
prefences, to automatically find the CW compromise, then you could
just let people vote the pairwise preferences that are most important
to them, the ones that they feel it's most important for them to vote,
either for strategy or for just supporting what they like. In other
words, use Approval.
.
Approval is the low-cost, count-efficient, and Condorcet-efficient
method. Instead of counting every pairwise preference, as CC-complying
rank methods do, you could just invite people to vote the pairwise
preferences that they choose to. That's an Approval election. We've
talked about the adantages of Score, whose easy, built-in fractional
ratings mitigate strategic misjudgements.
.
3. A compromise between those two choices would be Approval-IRV
(AIRV). Same as IRV, by my brief definition, but just let people rank
as many candidates as they want at any rank position (but especially
1st place).
.
For the purpose of my brief definition, it should be said that a
candidate tops a ranking if, uneliminated, s/he is at least one of the
candidates at that ranking's highest rank-level that has uneliminated
candidates.
.
Kevin Venzke pointed out that AIRV doesn't meet FBC. But it still
gives much Approval advantage, giving non-MM voters a way to elect the
CW without favorite-burial.
.
Its count is about as easy as ordinary iRV, but it's more
Condorcet-efficient. More count-laborious than Approval, but with
IRV's MMC compliance and no chicken dilemma.
.
I suggest that AIRV would be a great choice for a small group, needing
a relatively easily handcounted method, and wantng IRV's advantages,
along with better Condorcet efficiency, more considerate of non-MM
voters, nicer to majorities that aren't mutual majorities. In other
words, not as adversarial or uncompromising as IRV.
.
----------------------------------
.
As I said, Bucklin is strictlly for mutual majorities that are
entirely mutually trusting and trustworthy. if they aren't, then a
chicken dilemma will make Bucklin's MMC compliance quite meaningless.
.
When I say, "Bucklin" here, I mean "ER-Bucklin". It shouldn't even be
considered other than in the version that meets MMC. That version
incorporates a delay, to achieve that purpose. That's how it was
defined at electowiki, when I Iooked. We've discussed it here and at
EM before, so there's no need to repeat that definition.
.
But if you're going to propose and use a rank method, there's no
reason to use one that has the chicken dilemma. For example, why use
Bucklin instead of AIRV?
.
I'd suggest that, if you're wlling to have chicken dilemma, then make
it simpler and make the count easier, and just use Approval, or maybe
Score.
.
Michael Ossipoff