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

On 3 March 2017 at 22:05, Barak A. Pearlmutter <barak(at)pearlmutter(dot)net> wrote:
> In the Burlington election discussed, k=1 and n=3. That's about the
> simplest situation you can have, so any credible multiwinner system
> should perform flawlessly in that degenerate case, one would hope!
>

It is illogical to assume that methods and algorithms for one type of
problem, should apply to a different one, and expect similarly good
results.
Especially algorithms that are only defined for multi-winner cases,
and undefined for single-winner cases at all (reduce to a subset / a
different voting system).

I do not expect good multi-winner systems perform best for single-winner cases.

If there are two candidates who have exactly 50% of first & third
preference votes, and a candidate with 100% of second preference
votes, which winner should one choose for a single-winner case?

A multi-winner system can and sometimes do spit out randomly either of
these three as the first winner. Since the first two candidates are
equivalent, and the 100% second preference candidate is the most
consensus candidate.
Thus extrapolating single-winner results of multi-winner systems leads
to wrong conclusions.

This is kind of why I dislike single-winner elections a lot, as in the
above case all three candidates received meaningful and significant
amount of support - but actually electing any of those three can
results in huge social unrests and political backlash ("but everyone
liked candidate C", "but everyone preferred somebody else instead of
candidate C", "but B got the same support as A" (and vice versa))

--
Regards,

Dimitri.