[EM] Remember Toby (fsimmons at pcc.edu)

[Dave Ketchum]

On Jun 1, 2011, at 11:38 AM, Kathy Dopp wrote:

> I agree with everything you've said here re. simplicity etc.
>
> Condorcet with Approval to break Condorcet cycles would be great.
> Simple to explain, precinct-summable with the use of an NxN matrix,
> with N= # candidates and the matrix diagonal available for other data.
> (such as the total number of ballots cast or ?)

Sounds good until you think about Condorcet and Approval arguing as to  
what quality is worth ranking.  Approval wants ONLY desirable  
candidates; Condorcet can afford low ranks in case all those given  
higher ranks lose.

Note that each member of a Condorcet cycle has demonstrated CW ability  
vs every non-member.  Thus the cycle members are near to being tied,  
and properly compete among themselves for one to become CW.
>
> I like the idea of using Approval to count all except the last ballot
> position, whatever that would be. In the US, given current voting
> system capacities, that would be counting the first two ranked
> positions.

Attempted recovery - but the voter may, OR may not, have ranked one  
that would have been approved if the voter was thinking of Approval  
(and, the voter may have ranked only two).
>
>
> Upper margin error bounds could probably be calculated for each
> reported Condorcet matrix precinct tally so that selection weights and
> sample sizes could be calculated for post-election manual audits to
> publicly verify the accuracy of the reported election outcomes.
>
> Range voting would be too complex because it involves too much thought
> and strategizing for voters to determine how many relative points to
> give each candidate.

Agreed.

> Some of the other methods for resolving Condorcet
> cycles are too complex for most voters to understand and apply so that
> they can check the calculations.  IRV and STV methods are out, not
> only due to their nonmonotonicity, and their failure to solve the
> spoiler problem, but due to their fundamental unfair method of
> counting ballots which makes manual counting and thus auditing for
> election outcome accuracy virtually impossible.

Agreed.
>
> We ought to focus on how to make Condorcet/Approval voting
> understandable to the public and to election officials and show how it
> could be used with existing voting equipment, the existing problems
> with plurality it solves, etc.  I could work on developing the
> mathematics of post-election auditing sampling for it when I have
> time.

Not agreed - see above.
>
>
> Kathy
>
>> Date: Wed, 01 Jun 2011 02:46:20 +0000 (GMT)
>> From: fsimmons at pcc.edu
>
>> It seems to me that thevoters are more worried about the ballot  
>> type and ease of voting it than they are
>> of the exact counting rules.   There are several Condorcet methods  
>> that are clone proof and monotonic
>> without being too complicated.  I agree with Kevin that "elect the  
>> CW if there is one, else elect the
>> candidate ranked (or ranked above last) on the greatest number of  
>> ballots" is plenty simple, and is much
>> more satisfactory than MinMax or Copeland in other respects.
>>
>> But, as I said, what we really need to concentrate on is simplicity  
>> in votinig, i.e. how do we make ballots
>> that easy to use for "Hodge, fresh from the plough," as Lewis  
>> Carroll put it.
>>
>> It has been averred many times on this list that in Australia,  
>> where partial rankings are considered
>> spoiled ballots, the vast majority of voters fill out their ballots  
>> by copying "candidate cards" which are
>> published  sample ballots recommended by the various candidates.
>>
>> Asset voting makes this automatic for 100% of the voters.  That's  
>> probably going too far, so how do we
>> get a compromise between Asset voting and Condorcet?
> -- 
>
> Kathy Dopp