[EM] Bucklin variant, and in general

[Jameson Quinn]

2011/2/21 Kevin Venzke <stepjak at yahoo.fr>

> Hi Jameson,
>
> --- En date de : Lun 21.2.11, Jameson Quinn <jameson.quinn at gmail.com> a
> écrit :
> >If you knew your candidate was not A, though, you are guaranteed that your
> >second-place votes will count, perhaps against your first-place one.
>
> This is actually not different from Bucklin. Just replace "was not A"
> with "will not have majority on the first slot." The difference is that
> in Bucklin *every* voter has this problem.
>

I'm going to continue to insist on calling rated methods "MCA", as opposed
to the relatively-uninteresting and -unsupported ranked "Bucklin".

You're right that this same problem exists in MCA. I wasn't saying that
you'd created the truncation incentive, just that you'd failed to solve it.
Moreover, I find asymmetric guarantees to be worse than useless, because
they tend to weaken the "timeless decision theory" reason to cooperate.
(Google it if you haven't heard of it. The basic idea is that one should
cooperate in prisoner's-dilemma-like situations, even non-iterated ones, if
I believe my counterpart to be capable of understanding timeless decision
theory. I do not pretend that this is anything like rigorous in a
mathematical sense, but I do think that symmetry increases sympathy which
increases cooperation.)

>
> The strategic problem I see with my variant is that non-A voters will
> have incentive to bury A to get A voters to reveal preferences. But if
> the A's don't have any preferences to reveal, the strategy backfires.
> This is a bit similar to your concern that A will be unable to get second
> prefs from other factions. Factions declining to vote for A would have
> to believe that A voters have second preferences to give them.
>

Rationally speaking, you're right. But people are often deluded optimists.
It's relatively easy for several different factions to simultaneously each
believe that they themselves are the clear second frontrunner, and thus have
nothing to lose by truncating the first frontrunner.


>
> I am suspicious that if rated Bucklin (the best kind I think)


Call it MCA.


> lacked
> utility-like numbers on the ballot, it would not make sense. Here is an
> analogy: The polls are open all day. You can continually go back to the
> polls to add more preferences. If majority is reached, the election will
> end. Otherwise you will have to wait until the end of the day for the
> result. So, at 7AM I go in and cast my first preference(s). The question
> then is at *what time* should I add more preferences? At 9AM I have no
> additional information except that the method has not ended. Should I
> rely on pre-election polls for this? I.e. did we tell the media, I'll
> get nervous at 2PM and vote for another candidate? If so why did I pick
> *that* time? Wouldn't that just be based on what I heard other people say
> they were going to do, i.e.: we're making decisions in circles and it's
> not based on anything?
>

This is exactly why my preferred form of MCA is three ranks (preferred,
approved, unapproved). In your terms, you can only vote at 7AM or at 7PM.
More than two options provide only illusory benefits.

>
> I can just picture the poor fool who keeps adding votes throughout the
> day and then, when polls close, his original first choice is the winner.
> Should he be shocked?
>

That poor fool could legitimately wish that he'd at least added all his
votes at the end of the day. His sister, who voted her favorite at 7AM, the
winner at 8AM, and further votes throughout the day, could similarly wish
that she'd added her top two at 7AM and the others at the end. In other
words, everybody could wish that the system had been 3-rank. They still
can't be too regretful of their last-minute "insurance" votes, unless
multiple majorities were reached. And if that happens, the tiebreaker should
still be one that minimizes their regret by minimizing the importance of
second-choice votes. Choosing the pairwise champion (CW) is one such
tiebreaker; and using 2/1/0 Range-style totals would be close enough to that
to be an acceptable summable substitute.

Jameson
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20110222/1d780aae/attachment-0004.htm>