[EM] Ranking unacceptable candidates.

Michael Ossipoff email9648742 at gmail.com
Fri Oct 18 09:20:00 PDT 2013


First, another typo:

I said that, a difference in the two output-ranking is that, in
Beatpath, Libertarian shares 2nd place with Communist Party USA.

I meant that a diference in the two output-rankings is that, in
_Benham_, Liberarian shares 2nd place with Communist Party USA.

Juho--

I don't deny the desirability of the better expressivity and more
information gotten by encouraging complete rankings.

But, as you know, one must choose between different desiderata.

The problem is that, if completre ranking is encouraged, that can make
incentive for complete ranking, and that can make strategic need of
complete ranking. It can make a burial-war, as you yourself pointed
out.

Maybe, for public elections it's better give incentive for
unacceptable-truncated rankings, instead of for complete rankings.
Largely because it makes voting easier. That could be important.

For example, in my most recent poll at Condorcet Internet Voting
Service, there are 30 party platforms. No one wants to be
strategically-compelled to rank all of them. You don't want to need to
rank the 2nd worst over the worst, just so you can counter
Worst'voters doing the same thing to your favorite.

Michael Ossipoff





On Fri, Oct 18, 2013 at 1:28 AM, Juho Laatu <juho4880 at yahoo.co.uk> wrote:
> I think ranked methods should encourage full ranking. And I don't like active truncation in general. The key reason is that if people start truncating the worst candiates, and easily also their strongest competitors, then we start losing valuable information. People may also start to truncate those candidates whose supporters they expect to truncate their favourite candidates. In the worst case we end up in having a plurality style election instead of a ranked election (when people close to bullet vote, or rank only the candidates of their own opinion group).
>
> All information should thus be collected, and we should avoid opposite sides truncating each others. I guess it is quite common the elected candidate is disliked by many voters of the other opinion groups. But we still want to be able to elect one of the canddate in a rational way (= using preference information collected from the voters).
>
> Additional cutoffs/categories/approvals may help us in collecting full rankings from the voters. Harmless ranking is one approach to eliminating the risks of complete rankings. There can be also other approaches. In some elections we may allow the election to be inconclusive. Or sometimes we can afford to have multiple rounds (especially in small groups, that may also be more likely to use working strategies that normal people in large elections). The simplest rule could be to require the winner to be approved by majority. Otherwise more discussion is needed. Maybe the fifth election would not have any such rules, and people would have to accept the results anyway. Or maybe the required level of approvals would decrease in every election.
>
> My basic approach to large real life elections is that I want voters not to truncate, and to give sincere complete rankings (of all the potential winners at least). If they do, maybe we don't need any extra rules. Basic simple ranked methods would do. No reason to make strategy related additions to them if there are no problems. The methods can then be optimized to elect the best winner with sincere rankings. But if there are problems with truncation, then some kind of category based approaches would be a good way forward (instead of recommending truncation or other similar tricks that lead to loss of information).
>
> In the ranked elections that I have seen so far, strategic truncation has probably not been a big problem. Some people may truncate just because they are lazy, or because they still think in terms of some old voting method, or simply because they do not have any more preferences. We may learn more when we get more political competitive ranked elections (also ones that do not have the LNH property). In places where truncation becomes a problem we could use these diacussed tricks, like required approval level, power truncation, harmless ranking etc. In places where people generally rank sincerely in any case, different categories could collect additional useful information that need not influence the outcome of the election (e.g. messages to some candidates that they are liked although thy did not get eected, or messages to some candidates that 50% of the voters dinslike then strongly). There could be also elections where the strength and length of the mandate depends of how mu
>  ch "category based" support the winner got.
>
> Juho
>
>
> On 18.10.2013, at 4.10, Michael Ossipoff wrote:
>
>> That's strategically safe in IRV, because of IRV's Later-No-Harm compliance.
>>
>> But, with most rank methods, including Benham and Woodall, it would be
>> a strategically-bad idea, because it could increase the probability of
>> electing an unaccetable.
>>
>> You might say that, if you don't know which unacceptables are more
>> iikely to win as CW, then it's harmless to rank them. But, if you
>> think some unacceptables are worse than others, and if others are at
>> all likely to tend to share your preferences among them, then when
>> you're ranking U1 over U2, you're voting against a defeat of the one
>> (U1) who is more ilkely than the other U2) to win as CW. If there's a
>> tendency for that to be so, for the unacceptable candidate-pairs in
>> your ranking, then you're increasing the pobability that an
>> unacceptable will win--as compared to if you didn't rank any
>> unacceptabls.
>>
>> I guess that problem could be fixed by adding a count-rule saying
>> that, for any candidate you rank below your "unaceptable" mark,
>> ranking U1 over U2 shall not be counted toward keeping U2 from
>> pair-beating U1. But it is counted in determining if U1 pair-beat U2.
>>
>> If I refer to that again, I'll call it "harmless ranking". I only
>> suggest it for candidates whom you rank below your "unacceptable"
>> mark.
>>
>> Symmetrical ICT used "power truncation", whereby not ranking X  and Y
>> counts toward making X beat Y, and toward making Y beat X. SICT was
>> intended to be Appoval-like, for current-conditions, but maybe power
>> truncation wouldn't be desirable for Benham and Woodall.
>>
>> But harmless-ranking might be a good idea, for methods that don't meet
>> Later-No-Harm. Especially if it's known that some people will insist
>> on ranking candidates who are unacceptable to them.
>>
>> But it would also be fine to, instead, just warn voters that it's
>> better to not rank unacceptable candidates if you feel that your
>> higher-ranked ones are likely to be more popular with others, as they
>> might well be, if other voters are anything like you.
>>
>> I like the expressiveness of he balloting system suggested by Juho, in
>> which a voter can indicate a point above which candidates are
>> "Favorites" (even if they aren't all top-ranked, and a point below
>> which candidate are "Unacceptable", and a point below which candidates
>> are "Entirely, Abominably, Unaccepable".
>>
>> Since that would, without question, encourage people to rank
>> candidates they consider unacceptable, then it should be accompanied
>> by, at least, harmless-ranking.
>>
>> Michael Ossipoff
>> ----
>> Election-Methods mailing list - see http://electorama.com/em for list info
>
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info



More information about the Election-Methods mailing list