[EM] Anti-defection strategy device for IBIFA

[Michael Ossipoff]

Chris's Conditional Bucklin, and 3-Slot ICT are chicken-dilemma-proof, and,
with CD & FBC, they're methods that should be on any proposal-list.

But their shortcoming, for me, is that, other than beating C in the
chicken-dilemma example, you aren't really giving full general support &
protection to a candidate of a distrusted faction, when you protect against
defection.

3-Slot ICT leaves the demoted candidate without protection from
pairwise-count offensive strategy, or even innocent, nonstrategic
truncation, which, admittedly, might not happen.

Chris's conditional approval-rating disadvantages the demoted candidate by
not giving hir any top-votes. My conditional top-ranking avoids that
shortcoming.

It gives the demoted candidate a fully-effective top-vote, if s/he
qualifies by having more unconditional top-votes than any of the
unconditionally top-voted candidates on your ballot.

If the demoted candidate, the candidate of the distrusted faction, is in
your top-set or strong top-set, then you'd like to support hir as well as
possible, while protecting against defection by hir voters.

Also, my modified form of unconditional vote allows for a Conditional
Approval.

I'll call Chris's original Bucklin version "3-Slot Unconditional (top)
Bucklin". The "(top)" refers to the comparison of top-counts to determine
if a conditional vote is given.

I'll likewise refer to Chris's original conditional-vote option as
"conditional votes (top) option".

My conditional votes option, i'll call "conditional votes (unconditional)
option", because it compares unconditional votes.

Maybe I'd replace the words "top" and "unconditional" with the letters "t"
and "u".

My Conditional Approval doesn't need the "(u)", because it's the only
Conditional Approval.

Then, my Unconditional Bucklin is "Unconditional (u) Bucklin".

I should post complete and better wordings:

Conditional Approval:

Voters can approve as many candidates as they want to, by marking their
names on the ballot in a specified manner.

If desired, and approval of a candidate can be designated "conditional".

By a particular ballot, a conditional approval is given only if the
vote-receiving candidate has more uncondtional approvals than any candidate
unconditionally approved on that ballot.

The candidate with the most approvals wins.

Due to time considerations, I'd like to post this now, and then post a
complete wording of Unconditional (u) Bucklin in a subsquent post to this
thread.

Michael Ossipoff








On Fri, Nov 4, 2016 at 11:54 PM, Michael Ossipoff <email9648742 at gmail.com>
wrote:

> Alright, I guarantee that this will be brief, clear, and final:
>
> In fact, of course the "Unfavor" could be an automatic part of the
> conditional nature of the approval or top-ranking that you give to a
> candidate.
>
> To say it differently, the comparison could be based on how many
> unconditional approvals the compared candidates have, for the purpose of
> determining whether the conditional vote will be given.
>
> In that way, you aren't denying an approval or a top-ranking to the
> candidate to whom you give a conditional vote.
>
> And yet adding more unconditional approvals or top-rankings has no
> FBC-spoiling effect.
>
> So here's how I'd word the rule for determining whether a conditional vote
> is to be given:
>
> In Conditional Approval, or in Conditional Bucklin at the first rank:
>
> The conditional vote is given only if the vote-receiving candidate has
> more unconditional approvals (or unconditional top-rankings) than your
> unconditionally approved (or unconditionally top-ranked) candidate who is
> unconditionally approved (or unconditionally top-ranked) by the most
> ballots.
>
> In Conditional Bucklin, for conditional votes at ranks after 1st rank,
> it's just vote-totals that are compared.
>
> Michael Ossipoff
>
> On Fri, Nov 4, 2016 at 11:31 PM, Michael Ossipoff <email9648742 at gmail.com>
> wrote:
>
>> For that matter, as I was starting to suggest before, you could just
>> "Unfavor" the candidate of the distrusted faction (when approving hir).
>> Then, the determination of whether to give a conditional vote depends on a
>> comparison of hir "Favor" score* compared with that of your approved
>> candidate with highest Favor-score.
>>
>> .*A ballot increments a candidate's Favor-Score if it approves hir
>> without giving her an Unfavor.
>>
>> That's my best suggestion yet for this conditional Approval & Bucklin.
>>
>> In Bucklin, the top-ranking without an "Unfavor" would replace the
>> "Favoriteness designation" that I earlier today spoke of for Bucklin.
>>
>> But, like the Favoriteness designation, that's only for the top rank. For
>> subsequent ranks, Conditional Bucklin uses vote totals, as of just before
>> that round, just as I described earlier today.
>>
>> Michael Ossipoff
>>
>>
>>
>>
>> On Fri, Nov 4, 2016 at 11:08 PM, Michael Ossipoff <email9648742 at gmail.com
>> > wrote:
>>
>>> Well, if you treat your "Better Group" designations as absolute instead
>>> of relative, and if you don't demote one candidate when promoting another,
>>> then FBC won't be violated.
>>>
>>> It won't violate FBC for you unless you choose to make it do so.
>>>
>>> Spoken of in that way, it doesn't sound bad at all.
>>>
>>> Then, approving Favorite, and giving hir the "Better-Group" designation
>>> won't make Compromise lose, when s/he'd won before you approved and
>>> "Better-Group"ed Compromise.
>>>
>>> Obviously you should "Better-Group" all of your approved candidates
>>> except for the one that's a defection-threat.
>>>
>>> If people do that,then the "Better-Group" designation is a very good
>>> measure of faction-size.   ...just what is needed for the purpose.
>>>
>>> So maybe the Conditional Approval and the Conditional Bucklin that I
>>> defined tonight and t his afternoon would be pretty good after-all. Maybe
>>> better than comparing top-counts to determine whether to give the
>>> conditional vote, because of the stronger support that the A voters could
>>> give to candidate B.
>>>
>>> Michael Ossipoff
>>>
>>>
>>>
>>> On Fri, Nov 4, 2016 at 10:39 PM, Michael Ossipoff <
>>> email9648742 at gmail.com> wrote:
>>>
>>>>
>>>>
>>>> On Fri, Nov 4, 2016 at 9:45 PM, C.Benham <cbenham at adam.com.au> wrote:
>>>>
>>>> If you are only allowed to designate one candidate as your favourite, I
>>>> don't see how the method meets FBC.
>>>>
>>>> (endquote)
>>>>
>>>> By the fact that the favorite-designation has no effect or role in the
>>>> points sum.
>>>>
>>>> The favorite-designation doesn't give a vote or a point.
>>>>
>>>> If you approve Compromise, and Compromise wins, and then you decide to
>>>> approve Favorite too, that will in no way affect the matter of whether or
>>>> not Compromise outpolls Worst.
>>>>
>>>> Say Comprise was going to be your favorite, and you were going to
>>>> approve hir and designate hir as favorite. S/he would have won. Then
>>>> Favorite enters the race. So you'll approve hir too, and designate hir as
>>>> favorite,  instead of Compromise.
>>>>
>>>> Can that make Compromise lose, if Favorite doesn't win?
>>>>
>>>> All you're changing is the standard for whether a particular
>>>> conditional vote should be given.
>>>>
>>>>
>>>> Well alright, maybe, though you like Favorite better than Compromise
>>>> (that's why the arrival of Favorite changed your favorite from Favorite to
>>>> Compromise), maybe few other people do. Maybe Favorite is a  favorite to
>>>> fewer people. That can result in a conditional vote being given, when it
>>>> wouldn't have when Compromise was your designated favorite. The giving of
>>>> that other vote could cause Compromise to lose. Maybe Favorite isn't
>>>> winnable.
>>>>
>>>> So yes, you're quite right. I started this posting to tell you why you
>>>> aren't right, but you are right.
>>>>
>>>> And if, instead, you can designate as many favorites as you want to,
>>>> and you add Favorite to your favorite-designation list, instead of
>>>> replacing Compromise with Favorite, as favorite, then, the electorate-wide
>>>> favoriteness of your most electorate-wide favorite candidate will either
>>>> stay the same, or increase. So the addition of Favorite to that list won't
>>>> cause a conditional vote to be given that wouldn't have been given before.
>>>>
>>>> If your "favorite-designations" are absolute rather than relative,
>>>> you'll leave Compromise with hir favorite-designation. So maybe "favorite"
>>>> isn't the best word, because it implies unique best. Maybe "Better".
>>>>
>>>> If your "Better" designations are absolute instead of relative, then
>>>> the arrival of Favorite shouldn't make you un-designate Compromise, and so
>>>> you won't cause more conditional votes to be given by adding Favorite.
>>>>
>>>> ----------------------------------------------
>>>>
>>>> But here's a better possibility: Instead of giving that "Better"
>>>> designation to some, why not give a "Less good" designation to the
>>>> distrusted voters' candidate, the standard example's "Candidate B"?   .
>>>>
>>>> Then, instead of comparing which of the two candidates has more
>>>> "Favorite" or "Better" designations, compare instead which one has fewer
>>>> "Less Good" designations throughout the electorate?
>>>>
>>>> Adding Favorite won't have any effect on the number of Less-Good
>>>> designations received by your approved candidate with the fewest Less-Good
>>>> designations. You could reserve your Less-Good designations for candidates
>>>> of distrusted voters, like Candidate B.
>>>>
>>>> Or maybe it would be better to instead give a negative designation to
>>>> some candidates. ...a "Worse" designation to some of your unapproved
>>>> candidates. But then the comparison of the "Worse" totals would have less
>>>> relevance the matter of which of A or B is the more winnable, with the
>>>> larger faction.
>>>>
>>>> I guess I suspect that these same considerations apply to Conditional
>>>> Bucklin too.
>>>>
>>>> My reason for wanting something different from comparing the two
>>>> candidates' top-counts was that, by not rating B at top, you're giving hir
>>>> poorer protection against someone worse.
>>>>
>>>> I'd hoped that there was a way to avoid that, with a "Favorite"
>>>> designation that didn't assign a point or a vote. Evidently that runs into
>>>> a problem.
>>>>
>>>> So maybe, then, as with 3-Slot ICT, it's necessary to settle for poorer
>>>> protection of B, when you demote hir in order to protect against defection
>>>> by hir voters.
>>>>
>>>> If so, that means that this Conditional Approval can't be. And it means
>>>> that Conditional Bucklin must give poorer protection to Candidate B, when
>>>> protecting against defetion by hir voters.
>>>>
>>>> Regrettable. I'd hoped that something better was possible.
>>>> .
>>>> ------------------------------------------------------------
>>>> ---------------
>>>>
>>>> Well, recently, maybe last month, I proposed my own "Conditional
>>>> Approval" and "Conditional Bucklin". it made reciprocity the condition for
>>>> giving the conditional vote. I described a perfectly feasible
>>>> implementation for it.
>>>>
>>>> The problem was that it invites a strategy that results in another
>>>> chicken-dilemma--a sort of secondary chicken dilemma.
>>>>
>>>> But surely that's better than the original, primary, chicken dilemma.
>>>> It seemed a bit of a mess, but I don't think it failed FBC. But, then, I
>>>> just assumed that it didn't. Maybe I didn't look closely enough.
>>>>
>>>> By the way, I'd proposed that same conditional Approval & Bucklin some
>>>> years before too, at EM, and rejected it for the same reason.
>>>>
>>>> Maybe it has more promise than the one that I proposed tonight, but
>>>> maybe it still has too much problem.
>>>>
>>>> Maybe the kind of conditional comparison you suggested, the top-count,
>>>> is the only one that will work well enough.
>>>>
>>>> So maybe we have to settle for poorer help for Candidate B against the
>>>> candidates we like less, such as C. Maybe that demotion of B has to, to
>>>> some degree, abandon helping hir.
>>>>
>>>> That somewhat disappointed and pessimistic likelihood is my take at
>>>> this time.
>>>>
>>>> Michael Ossipoff.
>>>>
>>>>
>>>>
>>>> And with no mechanism for sometimes not counting normal below-top
>>>> approval the method fails Majority Favourite.
>>>>
>>>> So I judge this to be much worse than 3-slot  "Conditional" IBIFA or
>>>> MTA or MCA.
>>>>
>>>> Chris Benham
>>>>
>>>>
>>>> On 11/5/2016 7:35 AM, Michael Ossipoff wrote:
>>>>>
>>>>> Let me completely define Conditional Approval. This definition is
>>>>>> brief enough to be a proposal for a 1st reform from Plurality:
>>>>>>
>>>>>> Conditional Approval:
>>>>>>
>>>>>> You can approve as many or as few candidates as you want to, by
>>>>>> marking their names, on the ballot.
>>>>>>
>>>>>> You also have a place on the ballot where you can indicate your
>>>>>> favorite.
>>>>>>
>>>>>> (in a variation of these rules, you could indicate more than one
>>>>>> favorite if you want to.)
>>>>>>
>>>>>> For any approval that you give, you have the option of marking it as
>>>>>> "conditional".
>>>>>>
>>>>>> A conditional vote is given only if the vote-receiving candidate is
>>>>>> designated favorite on more ballots than is your ballot's
>>>>>> favorite-designated candidate.
>>>>>>
>>>>>> (...or (if ballots are allowed to designate more than one favorite)
>>>>>> on more ballots than is the candidate favorite-designated on your ballot
>>>>>> who is favorite-designated on the most ballots.)
>>>>>>
>>>>>> The winner is the candidate with the most approvals.
>>>>>>
>>>>>
>>>>> If you are only allowed to designate one candidate as your favourite,
>>>>> I don't see how the method meets FBC.
>>>>>
>>>>> And with no mechanism for sometimes not counting normal below-top
>>>>> approval the method fails Majority Favourite.
>>>>>
>>>>> So I judge this to be much worse than 3-slot  "Conditional" IBIFA or
>>>>> MTA or MCA.
>>>>>
>>>>> Chris Benham
>>>>>
>>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20161105/34dc7a99/attachment-0001.htm>