[EM] DSV

[Forest W Simmons]

Raphfrk had a question about Declared Strategy Voting.

Yes, originally DSV was limited to Batch and vote-by-vote versions of 
plurality strategies applied to ordinal ballots.

But on this election methods listserv we have expanded DSV to include 
any method that takes rankings or ratings and automatically applies 
some strategy to them in the voters' bests interests.  

Some of these are iterative, and some not.  For the most part the 
ballots are not converted into plurality form, because the early stages 
tend to have too much influence on which of many possible equilibria is 
chosen.

I like DSV methods that convert ordinal ballots into approval ballots, 
such as Rob LeGrand's ballot-by-ballot DSV with his strategy A.

However, ballot-by-ballot methods have too much of a stochastic flavor 
for most people, since they require random or pseudorandom ordering of 
the ballots. 

That brings us to batch DSV methods, which may or may not be iterative 
in nature.

Now here is a compromise that might be acceptable:

1.  All of the ballots are going to be converted into approval style 
ballots and the approval winner is elected.

2.  Anybody who wants to can vote an approval ballot.  Such a ballot 
requires no conversion.

3.  Anybody who wants to can choose a candidate to determine the 
approval cutoff on his/her ordinal ballot.

4.  Anybody who wants to can have his ballot converted to a copy of the 
approval ballot of a designated candidate.

5.  All other voters will submit ordinal ballots to be converted to 
approval ballots by machine.

In what order are these ballots processed?

First the ballots that require no conversion are processed.

These partial results (in the form of subtotals and correlation 
coefficients) are communicated to the candidates, who then specify 
their own approval cutoffs and the approval cutoffs for the voters who 
have delegated to them the right to do so.

Finally, the "machine" makes use of the statistics of all of these 
partial results to insert optimal approval cutoffs into the remaining 
ballots.

Note that this method is not iterative.  There are several steps, but 
no ballot gets converted to approval form more than once.

This method could be made iterative, by taking all of the approval 
statistics (including the machine converted ballots' statistics) and 
using that information to improve the machine generated cutoffs, etc.

But, for now, I am not proposing that.

Forest