[EM] 28 years of progress and a wakeup call

[Closed Limelike Curves]

It's a bit exaggerated, I agree. The thrust of his argument, though, is
correct. For tournament solutions like Schulze/Ranked Pairs, there are
strategic equilibria where voters all bury the hell out of each other and
the winner ends up being a complete lottery. DH3 is another example.

I definitely agree he's overstating his case when he says these methods are
useless. These systems have multiple strategic equilibria (especially with
winning votes). Often there are plausible equilibria around executing a
defensive strategy to protect the Condorcet winner. In small settings like
local elections, voters have less information, so zero-info honesty becomes
more important (a property score lacks).

I don't think Monroe applies to Condorcet//IRV, but I do think it applies
more weakly, in the sense that there are probably equilibria where the
wings engineer a center-squeeze to bury a Condorcet winner.

I'm unsure about Condorcet-Cardinal hybrids. My hesitancy comes from them
failing either FBC or later-no-help. I suspect this causes them to fail the
Condorcet criterion in some strategic equilibria as well, because
turkey-raising and favorite betrayal obscure the sincere Condorcet winner
in the polls. I agree this is a good avenue for further research, because
finding a method that elects the Condorcet winner in Myerson-Weber
equilibrium—but still satisfies strong sincerity in the zero-information
case—would be great.

But right now it looks to me as though, despite their mathematical
elegance, all the known Condorcet methods have plausible
strategic equilibria where the CW loses.

On Sun, May 26, 2024 at 6:14 AM Kevin Venzke <stepjak at yahoo.fr> wrote:

> Hi CLC,
>
> I think that WV Condorcet probably does get around "Monroe," at least
> given some
> assumptions about voter behavior that I think are realistic. I don't
> relish the
> idea of having to make that argument, but I think "Monroe" has limited
> scope. (It's
> actually hard to understand what the scope of the claim is meant to be.)
>
> The central problem with the methods Monroe investigates and then condemns
> is that
> the methods don't have a way to know which contests are important. It
> seems that
> voters opine on every pairwise contest, and if truncation exists at all
> it's just
> interpreted away, the same as split votes. Burial strategies and
> counter-strategies
> (which are also burial) will quickly confuse the scenario.
>
> I quote how he explains the situation:
>
> "All of these [paired comparison systems] ask for and use complex
> preference
> information. All of them then give the incentive for voters to exaggerate
> their
> preference differences over the most competitive alternatives up to the
> point where
> it is unclear who the main competitors are. At this point all alternatives
> can win,
> including alternatives that have absolutely nothing to recommend them
> except [that
> they are on the ballot]. All paired comparison systems violate
> [Non-election of
> Irrelevant Alternatives]; all of them are useless."
>
> Monroe realizes he could use first preferences to help. He observes that
> if a
> method satisfies majority favorite then it won't be as bad as Borda. But
> for the
> most part he doesn't seem interested in exploring nuances. His writing is
> a little
> hyperbolic.
>
> He clearly doesn't mean to say that all methods based on the pairwise
> matrix are
> bad, because his own proposal is still based on pairwise comparisons. It
> just
> blocks candidates from winning who don't have any first preferences.
>
> It's not at all demonstrated that you have to use first preferences as
> your cure;
> this is just his view of the evidence. And because there isn't a general
> proof
> regarding "paired comparison systems," it's not clear without further
> analysis that
> "Monroe" applies to Condorcet//IRV or Condorcet//Approval(implicit), let's
> say. And
> if it doesn't, that shows that there are avenues available to us.
>
> Kevin
> votingmethods.net
>
>
>
>
> Closed Limelike Curves <closed.limelike.curves at gmail.com> a écrit :
> > Why would Monroe (2001) not apply, unless you use the tied-at-the-top
> rule?
> >
> > On Tue, May 21, 2024 at 2:09 PM Michael Ossipoff <email9648742 at gmail.com>
> wrote:
> > > I meant that wv Condorcet can be relied on to elect the sincere CW,
> due to its
> > > excellent deterrence of offensive-strategy.
> > >
> > > On Tue, May 21, 2024 at 14:00 Michael Ossipoff <email9648742 at gmail.com>
> wrote:
> > >> What Meyerson & Weber demonstrated was that Approval’s Meyerson-Weber
> equilibrium
> > >> is at the voter-median. i.e. Approval homes in on where the CW is.
> > >>
> > >> But, additionally, it seems to me that Approval has chosen the CW on
> every EM poll.
> > >>
> > >> Other than wv Condorcet, Condorcet’s won’t always elect the sincere
> CW, because
> > >> of strategic-cycles.
> > >>
> > >> My & W also demonstrated that Plurality can keep on electing any pair
> of parties
> > >> forever at MW-equilibrium.
> > >>
> > >> …which of course is what it’s doing now. (with a bit of help from
> mass-media
> > >> promotion, & evidently at least occasional count-fraud.
> > >>
> > >> wv Condorcet will, because of how well it deters offensive-strategy.
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20240526/9fd6c0e4/attachment.htm>