[EM] Open letter to STAR voting promoters

[Closed Limelike Curves]

I believe we should expect that in the future there will be better
opinion polls that provide more information for governmental elections.
In other words, zero information elections will become rare -- if the
outcome is important, as it is in governmental elections.

I mostly agree, which is why I think the perfect information case is the
most important. However, in the perfect information case I think cardinal
voting (specifically approval) is the best you can do. Anything else is
going to create an incentive to be insincere, which can obscure the
Condorcet winner or the runner-up (the most important pieces of information
to have).

Say you know who the Condorcet winner and runner-up are, and you're working
in a monotonic system. Now, the best strategy is to rank the better one of
these candidates at the top, which minimizes the probability that they'll
be involved in a cycle. The other candidate should be ranked at the bottom,
to maximize the chances that someone else will pairwise-beat them: the more
candidates you rank above them, the more likely it is one will beat them.

Now we have the question of what happens to the other candidates. Ideally,
we want to minimize the amount of dishonesty. Perfect honesty is impossible
now: unless the top frontrunner is your sincere favorite, there's no way to
give a completely sincere ranking anymore. But it* is *possible to get
*semi-sincere
*rankings out of voters.

Here’s what I want to know: can we keep this property, but still get strict
sincerity for the zero-information case?

On Thu, Jun 6, 2024 at 9:22 PM Richard, the VoteFair guy <
electionmethods at votefair.org> wrote:

> On 6/4/2024 8:19 PM, Closed Limelike Curves wrote:
> < ...
>  > So, I suppose my real question is: is there any voting system
>  > that satisfies both properties? Sincere favorite and
>  > later-no-help are enough for voters who follow good polls to
>  > work out the Condorcet winner, by making min-maxing the
>  > optimal strategy. On the other hand, I don't know
>  > if they're necessary or if they're compatible with honesty in the
>  > zero-information case.
>
> Since you seem to have asked me this question, and no one else has
> answered it, I'll say ...
>
> I believe that attempting to "satisfy both properties" is likely to
> increase the failure rates of yet other fairness criteria.
>
> To me, a zero failure rate and a one-in-a-hundred-thousand failure rate
> are not significantly different.  Both are close enough to zero to not
> be a concern in elections.
>
>  > The place where this is bad is the zero-information setting. True
>  > zero-info almost never exists, but low-information settings crop up if
>  > you have a sincere Condorcet cycle, several clones, or a nonpartisan
>  > local election. In these settings, it would be very nice to have
>  > voters be perfectly honest.
>
> I believe we should expect that in the future there will be better
> opinion polls that provide more information for governmental elections.
> In other words, zero information elections will become rare -- if the
> outcome is important, as it is in governmental elections.
>
> I think honesty would be very nice in every setting.  But getting
> honesty requires using rules that don't motivate voters to vote
> dishonestly.
>
> In other words, if there's a way to cheat, cheating will occur.
>
> Richard Fobes
> the VoteFair guy
>
>
> On 6/4/2024 8:19 PM, Closed Limelike Curves wrote:
> >     * Every cardinal/rating method I know of is vulnerable to what used
> to
> >
> >     be called a "drama queen" meaning someone who exaggerates importance.
> >     Expressed another way ...
> >
> > I'm very interested in fleshing this out a bit more. In some situations,
> > this is the main drawback of score voting; min-maxing means your ballot
> > expresses less information than a full ranking and produces worse
> > results than most Condorcet methods. But in other situations, this is a
> > huge strength of rated voting that I think we really need to keep.
> >
> > The place where this is bad is the zero-information setting. True
> > zero-info almost never exists, but low-information settings crop up if
> > you have a sincere Condorcet cycle, several clones, or a nonpartisan
> > local election. In these settings, it would be very nice to have voters
> > be perfectly honest.
> >
> > The place where this is good is in the high-information setting where
> > voters follow the polls, or know . Then, every "sensible" voting method
> > will encourage exaggeration: if you rank or rate the best frontrunner at
> > the top, this maximizes your support for them, and therefore your impact
> > on the election. Similarly, ranking or rating a candidate last minimizes
> > your support for them.
> >
> > The only difference in the case of cardinal/rating methods is you can do
> > this without being forced to engage in order-reversal. This is a big
> > f*cking deal. Lots of voters will refuse to engage in decapitation
> > to protect a Condorcet winner (e.g. Begich in Alaska): it's
> > counterintuitive, makes voters feel icky, and reduces their favorite
> > candidate's shot at winning. Even if they /were/ willing, this kind of
> > dishonest order-reversal strategy can make it impossible to identify the
> > sincere Condorcet winner. ("Hmm, is RFK Jr. trailing because of
> > the brain worm or the major parties burying him?")
> >
> > In high-information situations, this kind of approval-thresholding is
> > the least-damaging kind of strategy:
> > 1. It's simple and obvious, so parties can't manipulate voters by lying.
> > 2. It has a single stable equilibrium point, so you don't need money or
> > party endorsements to prove you're a viable candidate.
> > 3. It's sincere, so it lets you identify the Condorcet-winner and the
> > Condorcet runner-up.
> > 4. It guarantees that if voters follow the polls, they'll choose the
> > Condorcet winner.
> >
> > So, I suppose my real question is: is there any voting system that
> > satisfies both properties? Sincere favorite and later-no-help are enough
> > for voters who follow good polls to work out the Condorcet winner, by
> > making min-maxing the optimal strategy. On the other hand, I don't know
> > if they're necessary or if they're compatible with honesty in the
> > zero-information case.
> ----
> Election-Methods mailing list - see https://electorama.com/em for list
> info
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20240607/8fe77a56/attachment.htm>