[EM] RP, Approval, MDDA, MDDAsc

[Michael Ossipoff]

Robert:

I didn't know about your November 25 (5 days ago) post until another
EM-participant told me about it.

I wasn't going to continue replying to you, but maybe I wasn't didn't
answer clearly enough. So I'll say it one more time, in this post.

I'd said:

*There's good consensus at EM about MDDAsc being the best deluxe
*
>* ranking-method.
*
not much conversation nor cross-conversation here to really discern a
"good consensus" about anything

[endquote]

No, actually there's been a fair amount of discussion about MDDA &
MDDAsc. There's never been _complete_ consensus at EM (only at
FairVote).

At one time there was wide agreement here for Condorcet(wv). And lots
of people here agree about Approval's merits, even when some don't
consider it their favorite.

But I don't think there's ever been as much consensus about a
chicken-dilemma-free FBC method.


.

You continued:


 i think, over the years, that there is consensus that IRV=bad

[endquote]

Not complete consensus. IRV is different in its tradeoffs, and not
favorite to most of us, including me. But I wouldn't call it "bad".
What's been bad is the dishonest promotion of it everywhere, denying,
attempting to conceal, its problem.



but Plurality=worse.
>* I'd suggest that those 4 methods should be offered in
*proposals for
>* municipal and state voting-system reform.
*
You reply:

it's dumb.

[endquote]

Not an argument that lends itself to an answer :^)

You're completely sure about what you believe, and you assert your
sureness loudly and ill-manneredly, often without supporting your
claims.

That's why I'd decided to not continuing to reply to you and why I've
begun filtering out your posts.

Why take the time to say anything to someone who is convinced that he
already has all the answers, and that everyone else is wrong? At least
you oould disagree politely. But no.

You continued:

especially when you include Bucklin on the list of 4.

  that's just dumb.

[endquote]


Bucklin meets FBC, gives good protection to top & middle ranked
candidates over unranked ones. ...and (by the option to skip
rating-levels) gives some protection to top over middle, and to upper
middle against lower middle.

There are lots of ways in which MDDA & MDDAsc are much better than
Bucklin. But Bucklin is still a good FBC method, with the advantage of
use-precedent.

Yes, Bucklin doesn't reliably elect the CWs. You fear an IRV-like
fiasco because of that. But there are important differences between
Bucklin & IRV:

What happened in Burlington happened because no one could protect the
CWs from elimination, unless the Repbublicans favorite-buried.

There wasn't anything that the Democrats could do to protect their
candidate from elimination. Favorite-burial was the only
CWs-protective strategy available to the Republicans.

Not so in Bucklin.

If the election had been by Bucklin, then the Dems, knowing that their
candidate is the middle candidate of the strong-3, and the likely CWs
(A middle candidate of 3 is the CWs if hir faction is bigger than the
difference between the extreme factions' sizes).

then they'd know that they should plump. ( & with additional strong
candidates, other voters would know that there's no need to rank past
the CWs).

And, if that isn't enough, the Republicans could, if they felt it
necessary, equal-top-rate the Dem. But they'd likely know that the
Dems would sufficiently protect their candidate by plumping.

So no, it isn't like IRV.

But it's still true that Bucklin might not always elect the CWs,
especially if people are more interested in protecting their entire
strong top-set, and therefore want to rank them all, even if it means
ranking past the expected CWs.

But then they have nothing to be angry about if the CWs dosn't win.

Nevertheless, one of the advantages of MDDA & MDDAsc over Bucklin is
that MDDA & MDDAsc, under sincere voting, always elect a majority CWs
(MCWs). It could be said that those 2 methods meet the Majority
Condorcet Criterion (MCC).

They always elect the MCWs, and they always elect the CWs unless
there's too much indifference. And if there's lots of indifference
about the CWs, then do you think that there would be lots of anger if
the CWs doesn't win? Whom could they blame but themselves?

In fact, for people who want the best candidate that they can get,
tactically-voted Approval would have done fine in Burlington.

The Dems, knowing that Middle is CWs if s/he's bigger than the
difference between the extreme candidates (and, if not, someone else
has a majority, and it dosn't matter how they vote), would plump for
their favorite. And the Dem would have won.

You might argue that the Repubs might not approve the Dem.

Someone probably would. Either it looks like the Progs are bigger than
the Repubs, or vice-versa. Whichever faction seems smaller than the
other would have tactical reason to approve the middle candidate.
Middle would win.

You might says that, in this election, Approval & Bucklin sound very
similar, and that Bucklin doesn't seem to be improving much on
Approval. Yes, but, with more candidates, Bucklin offers the
protections that I mentioned above.

Advantages of MDDA & MDDAsc over Bucklin:

Better avoids chicken-dilemma

Meets MCC (Majority Condorcet Criterion)

Is precinct-summable.

You're right about the desirability of precinct-summability. More
convenient for sure. Without precinct-summability, it's more difficult
(but not impossible) to prevent count-fraud.

(Public ballot imaging would prevent count fraud with any voting-system)



>* Let1 an initiative committee, &/or
*
>* the public (vial polls & focus-groups) choose among them, to choose the
*
>* best proposal for an initiative or referendum, or a legislated
*
>* voting-system reform.
*

you contact whomever you want in Burlington.  but you don't understand
when a "voting-system reform" backfires and is repealed.  it sets back
by a generation other "voting-system reform".

[endquote]

Of course. That's why you might want to reconsider proposing
RP(margins) in Burlington. If the people of Burlington are trusting
enough to believe you and enact it, then it will be the Rob Richie &
IRV fiasco all over again

RP(margins, or any Condorcet(margins) is completely vulnerable to
truncation, whether innocent or strategic. It would be a mess. You'd
be the next Rob Richie.

You continued:


1  Burlington and Vermont are pretty liberal places and we have
the most successful third party in the United States.  we even had a
gubernatorial election 2 years ago when the GOP got burned from
Plurality.  (the Libertarian candidate had 3x the margin between the
Dem Plurality winner and the GOP 2nd place.)  yet there is no talk of
voting system
reform since IRV was repealed in 2010 a year after it messed up.
no one's gonna be looking at your voting system reform based on
Approval voting.

[endquote]

Maybe, maybe not. Yeu don't speak for the general public. They can't
all be like you, or Burlington & Vermont would be bigoted redneck
places.

I've already acknowledged that there's psychological need for
rankings, and so, if people want rankings badly enough to want the
expense and trouble of setting them up, then rankings might very well
be best. But there are so many ways to count rankings that there might
not be any agreement on  a choice.

Loud, bigoted assertion & re-assertion won't help.



You continued:

it's because they will be skeptical to begin with.  they will ask
basic questions and you are unable to answer an extremely basic
question:  How can this relieve voters of the burden of tactical
voting when a voter cannot know, for sure, if their political
interests are best expressed by Approving their second choice or not?

[endquote]

I've answered that question several times. I've answered it every time
you've asked it.

I'll answer it one more time at the end of this post. For now, I'll
continue commenting on this post:

I'd said:

>>>* So rankings can soften voting errors, for rivals and for some
*
>>>* overcompromisers.
*
>>>
>>>* ...but not for all overcompromisers. Some overcompromisers are so
*
>>>* overcompromising that the only thing that can keep them from voting
*
>>>* Compromise over Favorite is if they have an opportunity to rank them both
*
>>>* at top, with the assurance that top-voting Favorite can't possibly hurt
*
>>>* Compromise.
*

this is totally speculative.  you have no idea who the
"overcompromisers" are, what their interests are, and what they would
do with whatever ballot design.  and obviating the "overcompromizer's"
motivation to over compromise by making a
ballot that encourages overcompromising (which is equally-approving
their second choice to their fav) does nothing with the other set of
voters that really want to support their favorite but still want to
express support for their second-choice over the candidate they hate.

[endquote]

You keep saying that you don't read my postings. Evidently not. But
please don't "reply" to what you don't read.

I've been saying that there are 2 kinds of overcomprisers:

1. The ones who'd be willing to rank Hillary below Jill, and would
thereby benefit from a rank method.

2. The ones who, if they had to rank without equal-rankiing & FBC,
would rank Compromise alone in 1st place.

That #2 strategy is common and well-known, and sometimes calleld
"exaggeration strategy", because it means ranking  the best of the
perceived frontrunners alone at top, and ranking the worse perceived
frontrunnerat bottom.

To help both of those 2 kinds of overcompromisers, you'd want a method
that allows ranking, allows equal top-rankng, and meets FBC (so that
the equal-top-ranking means something effective).

RP meets the 1st 2 of those 3 requirements, but not the 3rd. When an
overcompromiser wants to rank Compromise alone at top, you can't tell
that that isn't a good strategy for hir.

In Condorcet, ranking Compromise alone at top really is the best
strategy for someone who is convinced that the winner can only be
Compromise or Worst.

That's why FBC is desirable.

You don't seem to be getting that.

Aside from that, Approval-like voting, equal-top-ranking one's strong
top-set, is the way to maximize the probability of electing from your
strong top-set.

When a method meets FBC, and allows equal-top-ranking, and you choose
to rank your strong top-set in order of preference, then you're
increasing the probability of electing from your strong bottom-set.

So, to satisfy the psychological need for rankings, to accomodate both
kinds of overompromisers, it would be good for a method to allow
rankings, equal top ranking, and to meet FBC.

Condorcet isn't such a method, because it fails FBC.

I'd said:

>>>* ....in
*other words, for them, the method must allow equal
>>>* top-voting, and must meet FBC.
*
outside of any cycles (both with and without sincere voting),
Condorcet is fine avoiding favorite betrayal.

[endquote]

"Other than that, Mrs. Lincoln, how was the show?"

RP(margins) would be a truncation-cycle _mess_.

Any Condorcet method, and, in fact, any pairwise-count method, is
vulnerable to burial, unless defended against with good predictive
information. Any pairwise-count method could have the perpetual burial
fiasco (defined & described earlier), and the CWs wouldn't be winning.

More about that at the end of this post.

You continued:

  and i am convinced that cycles in reality will be very rare.

[endquote]

No. With RP(margins), truncation cycles will be fairly common.

With Condorcet(wv), there's some good deterrence against burial,
because the buriers don't have better predictive information than the
defenders do. But the need to guess the CWse (expected CWs) will be
there.

And, for the wing that doesn't like the CWs, and maybe dislikes hir
about as much as the opposite wing, there's little or no
burial-deterrence. Then, here comes the perpetual burial fiasco.

As I said, 'round about that time, you'd wish it was an Approval,
Score or Bucklin election.


You continued:

and that is because Nader voters would chose Gore over Bush

[endquote]


A lot of Nader, Bernie or Jill voters wouldn't rank Gore, Kerry, Obama
or Hillary. Partly for reasons of principle & aesthetics. Partly as
good strategy. In a pairwise-count method or Bucklin, never rank a
member of your strong bottom-set.

You continued:

  and Bush voters would choose Gore over Nader.

[endquote]

Yes.

(and it doesn't matter whether Gore voters would choose Nader over
Bush or not because Gore was the centrist candidate.)

[endquote]

With RP(margins), even if Gore were the CWs, truncation from the
progressive side could easily throw the election into a (sincere)
truncation-cycle, giving the election to Nader.(depending, of course,
on the faction-sizes)  Say, maybe RP(margins) would be a good idea
:^)



You continued:

it just resolves the election the same way as it would if the CW run
against any other candidate in a two-candidate
simple-majority race where everyone agrees how the votes should be
counted.  no reward for betraying your favorite

[endquote]

..because you believe that there'd never by a burial or truncation
cycle. If there could be one, then you could be sorry that you
top-ranked Favorite alongside Compromise.

You continued:

, all you do by bumping up your contingency candidate is help that
candidate defeat your fav. doesn't help you defeat your hated
candidate.

[endquote]

If you believe that there will never be a cycle, sincere or strategic

With RP(margins), expect plenty of cycles. Voting Favorite at top with
Compromise could then let Favorite beat compromise, allowing a
strategic or natural cycle to happen.

You continue:

>* Aside from that psychological
*need for rankings, Approval is better.
it's not merely a psychological need.  ranking is what we do to
discern and differentiate whom voters (with equally weighted
one-person-one-vote) prefer.

[endquote]

Yes. For polls in which there's known to be no strategic voting, I
prefer MAM, preferably without equal-ranking allowed.

But actual public official political elections are different.


in a situation in Burlington, often liberals will "approve" of both the Prog and
the Dem (over someone more conservative).

[endquote]

Maybe--I don't know what the Burlington Dems are like. Maybe they
aren't like Hillary, Obama, Kerry, Gore, etc.


You continued:

they'll take either the Prog or the Dem over the conservative, but
they may prefer one over the other.  Approval gives no mechanism for
that differentiation of "which liberal do you prefer?"

[endquote]

Above,in this reply, I told why tactical Approval voting would have
worked fine in Burlington, for people who want to elect the best
candidate they can get.

As for a better use of Approval, I'll get to that at the end of this post.

You continued:

 with such a voter (unless that voter does not
approve their second choice, in which case they cannot express their
preference of their contingency candidate or second choice over their
hated candidate.)

[endquote]

In Burlington, tactically-interested voters would approve the Dem from
one side or the other, if not from both sides.

You continued:

 >* Approval, Score, Bucklin & MDDAsc are reliable and solid.
*
You replied:

an opinion with no real support.

[endquote]

Admttedly MDDA & MDDAsc are pairwise-count methods. But they have
excellent chicken-dilemma protection, and that's the one thing that
can really justify pairwise-count.  ...and rankings, for that
matter--aside from the psychological need for rankings.

You continue:
 you cannot answer that basic challenge with either Approval or Score.

[endquote]

They'd have worked fine in Burlington. But their main use and value is
for a different kind of voting, not tactical, not for electing the
best individual candidate possible. More about that at the end of this
post.
,

You continued:

Bucklin is RCV and if it doesn't elect the Condorcet Winner, it has
the same problem of legitimacy that IRV has

[endquote]

No. I told why earlier in this post.

You continued:

 (which is why are we electing this Candidate B when more of us marked
our ballots that we prefer Candidate A?).  you'll never get past your
first 10 sentences introducing whatever MDDAsc to a legislature nor
voting public to adopt.  (it's also the reason i don't think Schulze
will get very
far, even if it's the best Condorcet.)

[endquote]

In Bucklin, as in Approval, if you want the best candidate you can
get, then you should try not to rank past the CWs.  ...just as you
should in Condorcet.

In general, in Bucklin, Condorcet, MDDA or MDDAsc, I'd advise people
to not rank past the CWse (expected CWs) if they're trying to elect
the best candidate they can get. Otherwise, I'd advise them to rank
only their entire strong top-set, or, absent one, to rank only the
candidates they really like.

...or, better yet, equal-top-rank your strong top-set, or the
candidates you really like, and forget about choosing among them.


Now, for a few things that I might not have said clearly enough before:

RP(margins): I've covered that well enough here.


RP(wv): I consider that the ideal best of the methods based entirely
on pairwise defeeats. MAM is the best method for sincerely-voted
polls.

RP says:

A defeat is affirmed if it's not the weakest defeat in a cycle whose
other defeats are affirmed.

[end of brief RP definition]

That makes RP, among the methods looking only at pairiwise-defeats,
the only one that never unnecessarily disregards a pairwise defeat.
The ideal pairwise-defeats-only method.

As I said, the clear best choice for sincerely-voted poBlls. MAM is
the RP(wv) version refined for best criterion-compliances even with a
small electorate.

The problem is that other properties are needed in serious public
political official elections.

...like FBC, and avoiding the chicken-dilemma problem. That's where
MDDA & MDDAsc come in.

All pairwise-count methods are potentially vulnerable to the
perpetual-burial fiasco. But, with MDDA & MDDAsc, you get something
important for usng pairwise-count: You get chicken-dilemma protection.

Approval:

You keep saying that I haven't answered your question about how
Approval avoids the problem of tactical voting.

Often, as in the Burlington example, tactical Approval voting would
work fine for people who want to get the best candidate they can.

But, aside from that, in general, there's a better way of using Approval.

Forget about trying for the best candidate you can get. There's
another kind of voting, another goal:

Approve (only) your entire strong top-set.

As I said, under the current (probably phony) political system, all of
the 99% would be so much better off with progressive, rather than
Republocrat, government, that those 99% rightly all have a strong
top-set (progressives) and a strong bottom-set (Republocrats and
anyone who is even worse).

But, if you don't have strong top & bottom sets (The 99% probably
won't have a winnable bottom-set under an authentic political system
such as would exist under better conditions, such as the Green
scenario).

then that's fine too. Under those conditions, with better media, and
honest elections, there'd be at least much more reliable predictive
information.

But, even if there weren't, that's fine too. Then, just approve the
candidates you really like, the very best.

Have trouble deciding where to draw the approval cutoff? No problem.
If you aren't sure where it should go, then it doesn't matter. There's
no wrong answer. Approve what you like best, however you might judge
that, by guess or feel.

If you want to discuss Approval, then you've got to get over your
belief that it's necessary to elect the best candidate you can get.

It's better to maximize the probability of electing from your strong
top-set, or at least electing someone you really like.

But evidently you do insist on electing the best candidate you can
get. Then Approval isn't for you.

You're exemplifying what I said about the psychological need for rankings.

Alright, but at least get a good ranking method. It isn't RP(margins).

Try MDDA or MDDAsc. Or Bucklin, if the proposal must be something with
use-precedent.

Michael Ossipoff
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20161130/3a8178c8/attachment-0001.htm>