[EM] When Voters Strategize, Approval Voting Elects Condorcet Winners but Condorcet Methods can Elect Condorcet Losers

[peter barath]

>In contrast, we show that Condorcet methods can elect
>a Condorcet Loser with non-zero probability when voters
>vote tactically. With strategic agents, approval voting
>is better at electing Condorcet Winners than Condorcet methods!

If we are fans of Condorcet

(and yes, some of us, understandably, are, being
Condorcet a natural extension of two-candidate-majority
voting, which is our everyday life, and if a candidate
beats another one, who would dare to say (s)he/it is not
the rightful winner, unless (s)he/it is beaten by
somebody/-thing else)

why don't propose a two-method system? If not for
other reason, it would be a good compromise between
Condorcetians and Approval/Rangeists.

The voters can give a Range-style ballot, but at first,
it can be counted as preference order (with equal positions
sometimes, but I propose 0-99 ranking to make almost equal
positions possible), and if there is no Condorcet winner,
Range counting follows.

Similarly, a preference order with an approval cutoff mark
can serve as Condorcet at first, and if not, as Approval.

The strongest argument against Condorcet, I mean the
only argument I'm capable to understand, is the dark horse:

many: Blue Red Hitler
a few: Hitler Red Blue
a great deal of: Red Blue Hitler

where Blueists easily tend to bury Red, voting Blue/Hitler/Red.
Do they this in order to make Blue into a Condorcet winner?
Of course, not. If Red doesn't lose the pairwise contest
against Blue, (s)he/it won't suffer it in buried state either.
So they bury in the hope of the next computation method
being in their favour.

If this next method is Approval or Range, Blueist will
still put Red down - just as they would do it in pure
Range/Approval voting. But they won't put her/him/it
under Hitler.

I can imagine this switch can hurt some fancy voting
criterium, but my best guess is it is worth being
still a Condorcet-method without the biggest problem
of Condorcet-methods.

Peter Barath

Digitális fényképezőgépek akár 10.000 Ft ajándék képelőhívással - FotoMarket Online Fotóáruház
http://ad.adverticum.net/b/cl,1,6022,99786,162268/click.prm