Re: Voting system R&D (Re: 2017 update to the SPI voting algorithm for Board elections)

On 2 March 2017 at 18:07, Barak A. Pearlmutter <barak(at)pearlmutter(dot)net> wrote:
> On 1 March 2017 at 13:47, Filipus Klutiero <chealer(at)gmail(dot)com> wrote:
>> I have received tens of mails from FVC and none discussed monotonicity or
>> any technical point.
>
>> This was not a comment on the substance of Barak's claim.
>
> In my discussion of these issues, I did my best to give pointers to
> grounded technical information that shows that STV and even its
> underlying IRV are poor voting systems, which actually exhibit major
> pathologies in practice. (E.g., electing the least-preferred of the
> top three mayoral candidates in Burlington Vermont; messing up when
> presented with actual Debian Project Leader ballots; leading to
> long-term two-party domination in the legislature using STV in
> Australia.)
>

SPI board elections are multi-winner.

All of the above are samples of single-winner.

We hold multi-winner elections by default, hence by definition it is
not IRV "instant-runoff voting" (alternative vote).

Multi-winner election means the three winners are equal, and it does
not matter who came first, second, or third.

I think it is a huge leap to conclude that particularities two-party
domination in Australia are due to STV, IMHO it is due to use of group
voting tickets. Also very offtopic.

STV is a voting system that strives to achieve proportional
representation, and that's a property which is desired for the SPI
board as it is a long-standing observation that the board is
disproportionally Debian member heavy.

For sake of clarity, we are working on moving SPI to PR-STV
"proportional representation through the single transferable vote".

> Ian and Joshua are dismissing these concerns, but have not given any
> technical grounds, either now or in the previous round of discussion.
>

The main reason to move away from condorcet vote, is that it currently
uses arbitrary elimination which results in skewed results for the
multi-winner case.
It typically leads to eliminating "consensus" candidates early and/or
arbitrary, whilst they could have gained a seat with PR-STV case.
("consensus" candidates are those that lost in all head to head
pairings (has little or no 1st choice preferences), yet are very high
2nd/3rd choice among all voters (people wouldn't mind this candidate))
(arbitrary in cases of A beats B, B beats C, C beats A with equal
weights as the top choice, one gets eliminated arbitrary without any
gains of subsequent preferences)

Extending mayoral candidates example, it is a contious design decision
to have the consensus candidate elected into the board. One could even
conclude, that your example of pathologies supports the case for SPI
to move to PR-STV =) luckily we are multi-winner elections, thus it
will not be the only board member.

> At the very least, it would seem to me prudent to craft a resolution
> which includes (a) some flexibility, so that the voting system can be
> changed more easily should there be reason to; and (b) which mandates
> making the full list of cast ballots public, so that pathologies in
> the elections can be detected.
>

Our current voting method is encoded with a board resolution.
We are updating it with a board resolution.

Thus this resolution doesn't change at all how voting systems are
changed or how elections are run. And this is intentional to keep the
scope of this resolution strictly about the algorithm alone.

Please stay on topic.

Please stop giving examples of single-winner elections, all of them
will be dismissed with prejudice as by default we optimizing for a
multi-winner election.

--
Regards,

Dimitri.