[EM] RWE, Participation

[Eric Gorr]

At 10:43 PM +0200 9/10/03, Kevin Venzke wrote:
>Eric,
>
>  --- Eric Gorr <eric at ericgorr.net> a écrit :
>>  If finding the option most acceptable to the group is of any
>>  importance to a voter, they cannot harm their own interests. A kind
>>  of person for whom this would not be important would be a dictator,
>>  someone involved in the overthrow of the government, etc...or
>>  basically anyone else who is unwilling to submit to the preference of
>>  the group as a whole.
>>
>>  As such, any method which is capable of finding the group preference
>>  would be one in which it is acceptable to fail Participation.
>
>I understand what you're saying, but there has to be a limit...  Basically,
>we should try to minimize the number of people who might throw a fit and
>regret the way they voted.  Maybe we can't be any more specific than that.

Let them throw a fit.

If the method is one that can reliably select the 
group preference, the fit has no merit and they 
(as well as everyone else) know this.

>It does seem to me that we've met most of what we need in Participation, if
>we meet weak FBC.  There should never be incentive to put B>A somewhere in
>your ranking if your sentiment is A>B.  If that is guaranteed, then the
>only strategy issue for the voter is when to misrepresent a sincere strict
>ranking as an equal ranking.  (That includes truncation.)

With a ranked method, I believe the only 
guarantee for such a thing is found in the 
general inability, if not impossibility, to 
gather accurate enough information to know when 
such insincere preferences would work. I believe 
in most of the examples I've seen, the winner was 
changed based on only a very small number of 
votes, when compared to the total, which would 
appear to be well within the margin of error of 
any poll that might be taken which could provide 
such accurate information to the voter.

Now, assuming an accurate enough poll could be taken....

Consider that before a poll was taken, in a close 
election, those candidates that were close would 
know that such insincere rankings could either 
help or harm them...considering that they cannot 
know which it would be before the poll was taken, 
it seems quite unlikely that they would be 
willing to take the risk. As a result, those 
supporting them would be informed to provide 
inaccurate information.

>I suspect it could be proven that a voter who uses only two ranks (due to
>equal ranking) in a method which never rewards order reversal, could never
>worsen the result from his perspective.

My intuition would say the same thing.

(assume the number of candidates > 3)

However, considering that insincere truncation 
will harm the ability for a method to find the 
group preference by shifting the decision on how 
at least two candidates compare with one another 
when that voter did have a sincere preference 
between them, suggesting that each voter only 
rank two candidates does not appear to be a good 
option.