[EM] Could a single specially-crafted new vote resolve all ties in a Shulze election?

[Scott Ritchie]

On 11/15/2011 04:29 PM, Juho Laatu wrote:
> On 15.11.2011, at 17.39, Scott Ritchie wrote:
> 
>> somewhat recently we had an election where there was a tie
>> between the 2nd and 3rd place candidates for an election where the top
>> two won.
> 
> I just note that if you use a (non-proportional) single-winner method
> that orders the candidates (maybe with ties) in the order of preference,
> then your results are not proportional. I.e. if 51% of the voters like A
> and B, and 49% like C and D, then you elect A and B. A proportional
> method would elect one of {A, B} and one of {C, D}. Maybe that is
> however what you want (two best candidates with no proportionality
> requirements). CIVS seems to support also a proportional mode (but maybe
> that is not what you want).
> 

Yes, PR is a separate issue (and one I do support in general), however
it's different enough from the current system that I can't just pick it
by fiat without running it by the council.


>> 2) Create a formal tie-breaking rule.  My intuition says that we can
>> give Mark Shuttleworth (who already has special privileges) a second
>> vote that he only uses in the case of a tie, add that vote to the box,
>> and then rerun the election.
> 
> Kristofer Munsterhjelm already noted that you could use that one vote
> (or multiple votes) directly to determine the preference order. That
> would be a good and simple approach.
> 
> I also note that if you want all the votes to be secret, you could allow
> also Mark Shuttleworth to cast a secret vote. He could give also an
> additional separate public vote that would be used to make the decision
> in case there is a tie. That vote could be required to rank all the
> candidates, so there is no need for further tie breaking. His public
> vote could just rank the candidates in a random order if he doesn't want
> to use his personal preference order. If he would give a random order,
> then we would already be quite close to just using a random order
> (lottery) to solve ties.
> 

It seems like CIVS is going to be missing the feature we need either
way.  Once an election is declared over, the results become public,
meaning that Mark would have to publicly break the tie.  I don't think
that's too much to ask of him, but it might be aggravating to have a
single finger to point (even though, logically, just about any voter or
nonvoter can be blamed when there's a tie).


Relatedly, if anyone wants my eternal gratitude and whatever marginal
eFame I can grant, I would really like a proper internet voting service
that solves the above problem.

If you're feeling really helpful, it would also be able to use Launchpad
OAuth to authenticate voters (currently we have to manually export a
large list of email addresses to CIVS, and then just hope the link in
the email is secure).

Thanks,
Scott Ritchie