[EM] Voter's Choice

[Forest Simmons]

In his masterful article at

http://www.barnsdle.demon.co.uk/vote/noagree.html

Mike Ossipoff proposes a clever meta-method called "voter's choice" to use
when there is no agreement among knowledgeable voters on which method to
use. I think Nobel Prizes have been given for less ingenuity. 


Each voter votes his preferences among the candidates with the
understanding that his preferences will be used (along with everyone
else's) in all of the common methods (based on preference lists) to
determine winners for all of the methods. (There will be a Borda winner,
an IRV winner, a plurality winner, as well as winners for various versions
of Condorcet, etc.) The voter also indicates on his ballot which one of
those (as yet undetermined) winners he would like his final (plurality) 
vote to go for. For example if he wants his final vote to go for the Borda
Count winner, and candidate A ends up winning the Borda Count, then his
final vote will go to candidate A. The candidate with the greatest number
of votes in this final plurality contest (not necessarily the plurality
winner in the list of method winners) is the grand winner.

This idea is very appealing to me, so I think it is worth extending to
include methods like CR and Approval (and now dyadic approval) that are
not based on (simple) preference lists. These other methods would be
accommodated if the voters were asked to rate each candidate on a scale of
zero to 100. And the preferences could be deduced from the order except
when two candidates received exactly the same rating.  In that case,
order could be imposed at random for the purposes of those methods that
require a complete preference list; after all that's what the voter would
have to do if he were required to show preferences where there was none. 

It's amusing to ponder whether any further advantage could be eked out by
allowing voters to rate each of the methods (as well as the candidates).

Well, don't let it give you a headache.

Forest