[EM] "true expressivities" of voting methods

[Richard Fobes]

On 5/27/2013 8:52 AM, Warren D Smith wrote:
> http://rangevoting.org/PuzzIgnoredInfo.html

Interesting.

Plurality and Approval collect so much less information that they do not 
noticeably ignore any information.

Instant-runoff voting obviously ignores information because it only 
considers preference information that "floats to the top".

Borda clearly does not ignore information, but it yields the wrong 
results -- unless somehow every voter separately ranks every choice.

When I was developing VoteFair ranking -- a.k.a. the Condorcet-Kemeny 
method -- I considered and then rejected the beatpath-like approach of 
looking at the biggest and smallest pairwise counts.  I rejected it 
partly because (similar to IRV) it ignores lots of the numbers (the ones 
that are not big or small).  (I also rejected it because it does not 
identify the second-most popular choice, the least-popular choice, etc.) 
  This concept of ignoring information is part of why I claim that the 
Condorcet-Kemeny method is better than the Condorcet-Schulze method. 
The opposite claim (that Schulze is better than Kemeny) tends to be 
based on counting the number (or importance) of fairness criteria that 
are met or failed.  When we finally measure how often those failures 
occur, the "information loss" of the Condorcet-Schulze method will 
become clear.  In contrast, the Condorcet-Kemeny method considers every 
pairwise count, not just the biggest and/or smallest pairwise counts.

Richard Fobes