[EM] Trying to have CD, protect strong top-set, and protect middle candidates too

[Michael Ossipoff]

Sorry to bother you yet again about this comparison, but I just want to say
one more thing about it:

It occurs to me that I was needlessly worried about a conflict, in MDDAsc,
between protecting against chicken-dilemma defection by B's voters, vs
protecting B from truncation & burial if s/he is CWs.

II realize that, in the standard chicken-dilemma example, C is
majority-disqualified, and is the only majority-disqualified candidate.

In that 3-candidate example, if B were the larger of {A,B}, and were CWs,
then, with the B voters not ranking or approving C, then, if the C voters
buried B under A, they'd thereby elect A as the only non-disqualified
candidate. (as several people have already pointed out).

So, in the chicken-dilemma situation, even if B were the CWs, s/he wouldn't
need an approval from A, for deterrence against truncation or burial. The
chicken-dilemma exists when people feel pretty sure that it exists. If it
looks & feels like a chicken-dilemma situaton, i probably is.  It's
disconcertingly obvious. And, then, there's no need to worry "What if it
might, instead, be a situation in which B needs approval from the A
voters?"

So my concern about that was needless.

It's clear when it's a chicken-dilemma situation, and MDDAsc works fine in
that situation. The voter has the option of dealing with the
chicken-dilemma situation when it's clear that there is one, and also
_fully, solidly, positively_ protecting, from truncation & burial,
middle-ranked candidates who need  that protection.

Michael Ossipoff





On Thu, Nov 24, 2016 at 10:27 PM, Michael Ossipoff <email9648742 at gmail.com>
wrote:

> Kevin suggested IC,MMPO, as a method that meets FBC, avoids
> chicken-dilemma, and has MMPO's (uncertain) kind of protection of a
> middle-ranked CWs.
>
> MMPOsc achieves that too,and the symmetrical completion avoids the
> Hitler-with-2-votes bad-example.  But the symmetrical completion that it
> needs spoils its truncation-proofness.
>
> So an advantage of IC,MMPO over MMPOsc is truncation proofness, because
> the IC avoids MMPO's bad-example, and so the symmetrical completion isn't
> needed.
>
> Forest's MDDAsc could be used with IC too. IC,MDDAsc would add a sort of
> Condorcet-Criterion compliance. I guess the symmetrical completion would
> still be needed, to keep Mono-Add-Plump,but, it doesn't do any harm to
> MDDA's solid, positive protection of a middle-ranked CWs to whom approval
> isn't denied..
>
> So that's four related methods:
>
> MDDAsc
> MMPOsc
> IC,MDDAsc
> IC,MMPO
>
> Some advantages among them:
>
> MDDAsc:
>
> Protection of a middle-ranked CWs is positive & solid, if approval isn't
> denied that candidate.
>
> MMPOsc:
>
> Chicken-dilemma deterrence is automatic. All middle-ranked candidates
> would receive some (uncertain) protection if they're CWs. But the
> protection of _any_ middle-ranked CWs is uncertain, and might or might not
> be there.
>
> The necessary symmetrical completion spoils the truncation-proofness that
> MMPO would otherwise have..
>
> Deterrence of burial and truncation depends on the would-be buriers
> knowing if the strength of pairwise oppositions favors them, or will result
> in penalty. It depends largely on whether the candidate under whom they
> bury the CWs has strong pairwise opposition other than by the CWs. Maybe
> that threat would depend on the voters of the defending wing, between the
> median and any possible burying-candidate,  purposely leaving hir only
> weakly pairwise-opposed (other than by the CWs). If there are a number of
> candidates under whom the CWs could be buried, and who could receive strong
> pairwise opposition, then it might be necessary to give low opposition
> (other than by the CWs)  to a number of them.
>
> Anyway, the protection of a middle-ranked CWs is always uncertain.
>
> IC,MDDAsc:
>
> Compiance with Improved Condorcet Criterion is added to MDDAsc's
> properties.
>
> IC,MMPO.
>
> Same as above. Also, because symmetrical completion isn't needed, due to
> IC, this method has, as an advantage over MMPOsc, that it's
> truncation-proof.
>
> -------------------------------------------------
>
> MDDAsc & IC,MDDAsc seem to win the comparison because the voter can give
> full, sold, positive protection to a candidate to whom s/he doesn't deny
> approval.  Chicken dilemma will be relatively uncommon, and so approval
> won't usually be denied.
>
> Michael Ossipoff
>
>
>
>
>
> On Wed, Nov 23, 2016 at 8:09 AM, C.Benham <cbenham at adam.com.au> wrote:
>
>> On 11/23/2016 10:59 AM, Kevin Venzke wrote:
>>
>> Hmm, with a movable cutoff MDDA already violates Plurality with three
>> candidates. Do you think symmetric-completion of the bottom can save it?
>>
>>
>> Yes.
>>
>> Something Toby Pereira wrote  (9 Nov. 2016) regarding "Irrelevant
>> Ballots" got me thinking. The criterion I defined just talks about ballots
>> that plump for nobody, but there can also be a problem with ballots that
>> only vote nobodies below, say, equal-top.
>>
>> Under MDDA with symmetric completion only at the bottom, adding  ballots
>> like that can wash away  "majority-defeat"
>> disqualifications and make the result less Condorcet-consistent.
>>
>> This has led me to think of a modification to fix this problem that looks
>> to be too good to be true, but so far I can't see how.
>>
>> *Voters submit rankings with an explicit approval cutoff. (I prefer
>> default placement to be just below candidates ranked below no-one).
>>
>> On the ballots that have been symmetrically completed at the bottom, find
>> the smallest set S of candidate/s that majority-strength pairwise
>> beat all the outside-S candidates.
>>
>> Disqualify the outside-S candidate/s and delete all the ballots that make
>> no distinction among the inside-S candidates.
>>
>> Repeat as many times as possible.  If at the end of this process more
>> than one candidate hasn't been disqualified, elect the
>> one of those by normal MDDA (SC).*
>>
>> I am fearful that this might fail FBC  and/or mono-raise, but I can't
>> (yet) see how it does.
>>
>> 45: A>B>>C
>> 10: A=B
>> 40: B
>> 05: C
>>
>> 100  ballots. After symmetrically completing at the bottom we get  A>B
>> 47.5 - 42.5,    A>C  75-25,    B>C  95-5.
>>
>> Normal MDDA(SC)  disqualifies only C and then elects the most approved
>> candidate, B.
>>
>> My suggested version first disqualifies C and then deletes the 5 C  and
>> 10 A=B  ballots and then disqualifies B leaving  A (the CW)
>> as the winner.
>>
>> What do you think?
>>
>> Chris Benham
>>
>>
>> On 11/23/2016 10:59 AM, Kevin Venzke wrote:
>>
>> Hi Chris,
>>
>>
>> ------------------------------
>> *De :* C.Benham <cbenham at adam.com.au> <cbenham at adam.com.au>
>> *À :* Michael Ossipoff <email9648742 at gmail.com> <email9648742 at gmail.com>
>> *Cc :* EM <election-methods at lists.electorama.com>
>> <election-methods at lists.electorama.com>; Forest Simmons
>> <fsimmons at pcc.edu> <fsimmons at pcc.edu>
>> *Envoyé le :* Mardi 22 novembre 2016 10h51
>> *Objet :* Re: [EM] Trying to have CD, protect strong top-set, and
>> protect middle candidates too
>>
>> On 11/22/2016 9:25 AM, Michael Ossipoff wrote:
>>
>> >>With MDDTR, if your plump for X makes hir lose, it's because you added
>> a ballot. It has nothing whatsoever to do with the fact that the new ballot
>> plumped for X.
>> >>Your ballot made X lose in spite of the fact that it was a plump for X,
>> not because it was a plump for X.
>> >>But in IRV, when you make X lose by raising hir from last place to 1st
>> place, that raising of X was the only thing that you did, and it is the
>> reason why X lost.
>> >
>> >That "distinction" is meaningless and completely useless.  The idea that
>> adding a ballot is "something you did" that rates a mention is ridiculous.
>>
>> I'm not sure about this specific example but I think this kind of
>> distinction could be a useful defense. For IRV I might argue that a
>> mono-raise failure happens not just from raising the winner but also
>> *lowering* some other, incidental candidate. The reason mono-raise
>> failures are offensive is that supposedly the voter has done nothing but
>> aid the preexisting winner. But at least in IRV it is not so clear as
>> that.
>>
>> >Regarding MDDA,  symmetrically completing the ballots only at the bottom
>> and having a moveable approval cutoff fixes its failures of Mono-add-Plump
>> >and Plurality and Irrelevant Ballots Independence and in my opinion
>> makes it a good/acceptable method.
>>
>> Hmm, with a movable cutoff MDDA already violates Plurality with three
>> candidates. Do you think symmetric-completion of the bottom can save it?
>>
>> For whatever interest it may be, I calculated the "DNA" for the method
>> you describe and got the exact same 343-digit code as for ICA(explicit).
>> That's the first time I've hit a method I already had...
>>
>> Chris Benham
>>
>> No virus found in this message.
>> Checked by AVG - www.avg.com
>> Version: 2016.0.7924 / Virus Database: 4664/13460 - Release Date: 11/22/16
>>
>>
>>
>> ----
>> Election-Methods mailing list - see http://electorama.com/em for list
>> info
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20161125/48b20217/attachment-0001.htm>