[EM] Remember Toby

[robert bristow-johnson]

On Jun 3, 2011, at 10:30 PM, Dave Ketchum wrote:

> Been a busy day on this thread.  I will try for Condorcet, to sell  
> that it can be good and usable by voters without requiring much new  
> understanding by them.
>

my sell for Condorcet compliant is more of a negative: If you don't  
wanna elect a Candidate B when more voters prefer Candidate A, you use  
a Condorcet method.  if they say "Hmmmm. I think this Condorcet is  
sorta tricksy", i yell back "Why do you want to elect a candidate when  
that candidate is rejected by voters in favor of another candidate?"

Condorcet is simple.  the Ranked Ballot is simple.  all it says is who  
you vote for when any two candidates are drawn, if you chose to select  
between the two.  any two-candidate comparison can be made and every  
ballot counts equally.  all candidates start out as potential winners,  
if a candidate is beaten in any paired runoff, he/she is marked as a  
loser.  the candidate who is left standing (not a loser) is the winner.

of course, this leaves off the deficit of the potential cycle.  that's  
when us Condorcet proponents get to appeal to Arrow's Theorem (then  
you hope some sophisticated voters that were involved in the IRV  
debate nod their heads).  where Condorcet fails Arrow is because,  
although each ballot is linear in ranking (there are no ambiguities  
that we know how the voter prefers any candidate to any other), the  
results from

> Ranking:
>      IRV ranking, learned by many, is a start, with equal ranking a  
> trivial addition.
>      Approval voting permissible and usable as valid Condorcet by  
> using a single rank number.
>
> How many rank numbers?  Three, as in IRV, is  probable reasonable  
> minimum.

here in Burlington, we had five levels, and there were five candidates  
(plus Write-In) on the ballot in 2009 and a similar number in 2006.   
no truncation was forced.

i think the number of levels has to be limited in the rules because of  
real estate on the ballot.  and i think that ballot access laws should  
be tough enough that the number of candidates often is (say, within  
90% of the occurrences) equal to or less than the number of levels.   
if it begins to appear that the number of candidates exceeds the  
number of levels regularly, legislatures should notice and increase  
the ballot requirement (number of signatures needed to get a candidate  
on the ballot).

>  More needs thought, but not necessarily many - usability of equal  
> ranks minimizes true need for more.

certainly agree with that.  that was only a problem for IRV (i guess  
it would be a problem for Borda, if these total points needed to  
remain integer valued).  i don't consider "usable for IRV or Borda" to  
be a particularly valuable property of a voting system.  i think  
Approval requires more thinking from the voter and i think Score does  
also.  and i don't like at all this Asset system thems trying to foist  
upon us (with its smoke-filled rooms and all).  the ranked ballot  
requires only for a voter to decide between any two voters just as  
they would if it were only those two.  and Condorcet counts it  
precisely "one person, one vote", just as a two-candidate simple- 
majority election counts it.  what the voters have to accept is that  
they have to decide about *every* candidate, not just their favorite,  
by Election day.  why is that too much to ask?  (we normally require  
voters to make up their minds about the content of an election by  
Election Day.)

i guess i'm still unmoved from using a ranked ballot with sufficient  
number of levels to accommodate every voter's expression of  
preference, and using Condorcet to decide the result.  which Condorcet- 
compliant method is something i'm more agnostic about.

--

r b-j                  rbj at audioimagination.com

"Imagination is more important than knowledge."