[EM] Friendly Approval

[Rob Lanphier]

Thanks for the reminder, Forest!  I forgot about Toby Nixon's
Condorcet mailing list and the flurry of activity around that time.

I looked up Toby's website, and found his Twitter:
https://twitter.com/tobylnixon

...and I tweeted about the possible state of his ears:
https://twitter.com/robla/status/1584810571072880641

As I recall, Yahoo was in the process of shutting down the archives
for the groups.yahoo.com back in 2019 (when I was at Internet
Archive).  I didn't think to ask someon to back up that mailing list,
but it seems that the cover page was backed up in 2013:
https://web.archive.org/web/20130121034857/http://groups.yahoo.com/group/Condorcet/

...and there may be some raw .MBOX files that a determined individual
could collect:
https://datahorde.org/how-to-recover-your-yahoo-groups-from-the-internet-archive/

Also, it looks like Jeffry Fisher backed up a few messages on
groups.io.  I see one of the messages that I sent back in 2005:
https://groups.io/g/CondorcetVoting/message/29

I guess I was still thinking that Schulze would be just fine back then.

Y'all may be able to reconstruct the contents of the mailing list from
your respective personal mail archives.  I might too.  I'm too lazy
though.  Someone else is going to have to do it.

I don't remember what "IRRV" was, so I looked it up on electowiki and
found this:
https://electowiki.org/wiki/IRRV

I should have known that "RR" was "round robin", but the method didn't
take the world by storm (though I suppose there's "Ranked Robin" now
occupying the "RR" abbreviation).  I suppose "railroad" is what many
people think of when they see "RR" (certainly, I do).

Anyway, I put up an electowiki page for Toby Nixon:
https://electowiki.org/wiki/Toby_Nixon

Given that he's notable enough for Wikipedia, he's notable enough for
electowiki.

Rob



Rob

On Mon, Oct 24, 2022 at 6:36 PM Forest Simmons
<forest.simmons21 at gmail.com> wrote:
>
> Remember Toby Nixon, a Washington State legislator who seriously consulted with the EM List members to find the simplest decent Condorcet proposal with a good chance of being adopted?
>
> The method we settled on was Approval based Benham, which we called DMC for Democratic Majority Choice (or Definite Majority Choice for Republicans).
>
> The Benham elimination formulation was the first of several equivalent descriptions: list the candidates in approval order and eliminate candidates one by one from the bottom of the list, until some candidate becomes undefeated among the remaining.
>
> Jobst was the first to point out that Ranked Pairs, Beatpath CSSD, and River were all equivalent to this method when defeat strength was gauged by winning approval.
>
> We came "that close" to getting DMC adopted, before some spoil sport scuttled the project.
>
> Another simple approval based Condorcet efficient method (even nicer than DMC in some ways) is ASM, Approval Sorted Margins: sort the approval order into a beatpath by transposing adjacent out-of-order-pairwise candidates, giving priority to the pairs with smallest approval margins.
>
> Another excellent one is approval based SPE (Sequential Pairwise Elimination).
>
> Another one is Approval Chain Climbing.
>
> And finally approval based top two runoff, a truncated form of approval based SPE, that we only mention because its score based STAR version has gained some traction recently.
>
> The only drawback is that these methods, without exception, in their present forms have not been successfully adapted to Ranked Choice Voting style ordinal preference ballots.
>
> The closest thing to it has been RCV style ballots with optional approval cutoffs, still very unsatisfactory for voters uncomfortable with the decision of exactly where they should put that cutoff for optimal effect.
>
> But now, finally we have a well-defined notion of approval in the Universal Domain category (i.e. restricted to using only ordinal information from the ballots) based on Kristofer's idea of "friendly" candidates. [This is my adaptation of his idea.]
>
> A candidate X is friendly to candidate Y if it does not directly defeat Y pairwise.
>
> And by extension we say that a ballot is friendly to Y if B's top ranked candidate is friendly to Y.
>
> We define candidate Y's friendly approval to be the number of ballots that are friendly to Y.
>
> Another way to say this is the total first place support of the candidates that are friendly to Y.
>
> This friendly version of approval suddenly opens up all of the wonderful approval based methods mentioned above (and more) for inclusion in the Universal Domain category!
>
> Furthermore, it automatically confers Landau efficiency onto most of them, because if  X covers Y, then X's friendly approval cannot be less than Y's.
>
> Thus DMC based on friendly approval is Landau efficient.
>
> We can expect Friendly Approval to have high Voter Satisfaction Efficiency, because of the essential role of first place votes in its definition.
>
> So now what's keeping us from proposing Friendly Approval based DMC?
>
> -Forest