Kevin,
I just realised that my suggested IIB-fix of MCA does cost a criterion
compliance:
Later-no-Help. Adding middle-ratings can help top-rated candidates by
maybe
increasing the Max Pairwise Opposition of their rivals.
I consider having LNHelp and LNHarm in (at least probabilistic)
balance to be more
desirable than either by itself, so I don't mind losing MCA's
LNHelp (since it badly
fails LNHarm). But I have to withdraw my suggestion that MCA doesn't
have (for a
3-slot method) a maximal set of properties.
And I think there are better 3-slot FBC-complying, LNHelp failing
methods that use
MPO information combined with ratings information (than my suggested
modified MCA).
One possibility: "If any candidates have a top-ratings score not
smaller than their MPO
score, disqualify the other candidates. Elect the undisqualified
candidate with the highest
Approval minus MPO score".
Chris Benham
Kevin Venzke wrote:
Chris,
--- Kevin Venzke <[EMAIL PROTECTED]> a écrit :
Kevin,
Kevin Venzke wrote:
As far as my strategy simulation is concerned, this rule change raises
the
question of how voters should evaluate the possibility that they elevate
a
candidate to the top spot on first preferences only to see him lose due
to
pairwise opposition.
Chris replies:
I don't fully understand this point. Any candidate who would win in the
first round of regular MCA would
also win in the first round of my suggested version, and in both the FPW
can win in the second round.
The only difference is that my version is more likely to have a
first-round winner, which I suppose in the
FBC-complying 3-slot ballot version might be a bit self-defeating. In
your FPP-approval ballot version
I don't see how it greatly complicates the strategy.
Currently the value of a first-preference vote for A is estimated as the
likelihood that A can achieve majority times the likelihood that no
candidate will achieve majority (e.g. if a majority is guaranteed then no
vote is of value) times the difference between A's utility and your
expectation should the election be resolved on approval.
With your rule you no longer simply break ties between one candidate's
majority and "no majority"; you have to compare against each other
candidate FPP-style. And you can't simply compare the candidate's utility
to the approval expectation, because the candidate could lose despite
coming in first.
If I were implementing this method I would probably have voters keep track
of their expectation when each candidate is the TRW but has too high
pairwise opposition. This kind of approach so far has produced a lot of
intelligent behavior. It has a couple of downsides though: 1. Voters can't
predict the value of situations which weren't observed to occur in the
polls, and thus won't try to create them, and 2. There seem to be a number
of "cum hoc ergo propter hoc" mistakes where voters vote for situations
that have coincided with outcomes they liked, but which didn't necessarily
cause them.
Kevin Venzke
_____________________________________________________________________________
Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail http://mail.yahoo.fr
----
Election-Methods mailing list - see http://electorama.com/em for list info
----
Election-Methods mailing list - see http://electorama.com/em for list info
