On 4-Dec-08, at 2:33 PM, Chris Hostetter wrote:
: Being the likely two candidates for winning. My guess is that
: narrowing to the two most popular options first would make #2 the
: winner, while voting on the top 10 (w/o any strategy for winning)
: would make #1 the winner.
limiting to only voting for the top 2 seems unrepresentative since
more
then one apache_solr_c_red.jpg variant tied for 2nd.
: fun, fun. So people who want one of these options to win should
vote
: only for that option, really.
Perhaps instead of just ranking top 5, we should ask committers to
rank all of the choices on the final ballot to eliminate the
"strategy" factor you are refering to ... i think we can trust all
committers to understand this, but if someone botches it (or refuses?)
we'll just shift the number of points each item earns down by the
appropraite number (so if you want your 1st rank to earn 10
points, you must list all 10, if you only list 4 then your top
ranked item
only earns 4 points)
Eliminating strategic voting merely biases the outcome toward the logo
without the vote splitting problem. That is no solution.
It is better to allow strategic voting, as that is the only way for
voters to express certain preferences in this system.
I would personally prefer more of an "elimination"-style vote (i.e.,
STV). Each voter lists the logos they prefer, in order. The logos
are ranked by first place votes. The last in the rank is eliminated
from the contest, and anyone who had that logo as their first-place
vote has their vote transferred to the next logo on the list, if any.
Iterate until two logos remain. There is no danger of vote-splitting
and the outcome maximizes global welfare in terms of binary
preferences (well, probably not, due to Arrow's theorem, but it does a
good job regardless).
-Mike
