[EM] it's been pretty quiet around here...

[Dave Ketchum]

On Aug 14, 2010, at 2:18 PM, robert bristow-johnson wrote:
> On Aug 14, 2010, at 5:37 AM, Kristofer Munsterhjelm wrote:
>>
>> Since FV thinks IRV is so nice, it's to their benefit to link  
>> preferential voting, the concept, to IRV, the method, so that  
>> others thing "oh, either IRV or Plurality". Since IRV appears  
>> better than Plurality (at least until the summability issues are  
>> encountered), this makes it relatively easy to slip in IRV, and the  
>> theory then goes, to go from IRV to STV, which is much better.
>>
>> It doesn't appear that we can change FV's minds from IRV to  
>> something better (like Condorcet). When you dig really far down,  
>> the issue boils down to "weak centrist! Condorcet winner! weak  
>> centrist! Condorcet winner!" and there you go -- and then they  
>> sprinkle LNHarm and *perhaps* burial resistance on top.
>
> my experience with Rob Ritchie is that IRV is the only method with  
> an ice cube's chance in hell of being adopted in a governmental  
> election.  the claim is that IRV can be directly related to the  
> traditional delayed runoff and that it is no different, except for  
> no delay (which has the measurable difference in that many more  
> voters participate in the instant runoffs than in the delayed  
> runoff).  but, for that to be true, it should have no more than 2  
> rounds with the top two of the first round going into the second and  
> final round.  of course, that doesn't fix the problems demonstrated  
> in the 2009 Burlington mayoral election (because the "true majority"  
> winner would not have made it to the runoff in either case).

Perhaps we will do better if we aim at attacking their weaknesses.

Delayed runoffs were invented to attack an experienced Plurality  
weakness - its voters cannot fully express their desires.  The French  
had a major experience of this in 2002 - Le Pen, a minor candidate,  
got to the runoff in place of the deserving winner, and lost as  
deserved.  We should not talk of runoffs unless we are prepared to do  
better.

IRV's "instant runoffs" can fail much as the Plurality failure I  
mention above.  Here the voters can more completely express their  
desires.  Trouble is, IRV has a formula for ignoring parts of the  
ballot data, with results such as were demonstrated in Burlington in  
2009.

In Condorcet the voter ranks candidates as in IRV.  Difference is that  
the counters read all that is voted, scoring a mini-race between each  
pair of candidates.  The best candidates will win the election via  
winning all of these races.  Else the best candidates will dispose of  
those weaker but require further analysis to decide on a winner.
>
>> Cardinal ratings technically pass both because it can pass IIA  
>> since it doesn't care about universal domain. However, I think that  
>> CR (Range, Score, etc) will be hard to get passed, since it doesn't  
>> even pass Majority. Even if Warren is right and social utility  
>> comparisons are better than majority rule, most people associate  
>> democratic fairness with that if some candidate is preferred by a  
>> majority, he should win. There are also the tactical issues: CR  
>> reduces to Approval (as Youtube raters found out)
>
> and Approval can reduce to Plurality bringing along the same  
> strategy problems of Plurality.
>
>> and pretty soon voters who want their vote to count must haul  
>> around concepts like "maybe frontrunner, plus" (LeGrand's Approval  
>> strategy A), something which really should be inside the method  
>> rather than outside.
>>
>> Thus we can't follow FV; and while we could advocate cardinal  
>> ratings, I don't think that would be very successful (and in any  
>> event, should be DSV instead). That leaves Condorcet, and so I  
>> think there should be an organization or group or at least some  
>> sort of coherent support for Condorcet.
>
> well, there used to be a condorcet.org or condercet.com (neither  
> have an active web page, although the .com has a dumb page put up by  
> the registrar, just like my audioimagination.com does).

Seems like we need this.
>
>> (Alas, I'm not a very good organizer and I'm about 5000 km away.)
>>
>> What should such a group do? First, it should state that the  
>> concept of ranked voting is different from what method may be used  
>> as its back-end. Second, it should have a clear and easily  
>> understandable name for Condorcet, or for the Condorcet method it  
>> settles upon. The former could be done more simply: "round robin  
>> voting", "maximum majority voting", "championship" or "tournament"  
>> voting (but beware of equating it with an elimination tournament),  
>> etc.
>
> Warren has used the term "beats-all winner" for the Condorcet winner.
>
Hope we can do better.
>
>> The latter would be more difficult, as Schulze, for instance, is  
>> hard to explain.
>>
>> For reasoning, it might point out that if you put all the voters on  
>> a line, and cancel out the leftmost with the rightmost until one  
>> voter remains, the candidate closest to that voter wins -- if  
>> that's not too advanced.
>> It might also show that if there's a CW, no recall by any of the  
>> other candidates can work against him, because a majority prefers  
>> him to each of the other candidates. That particular argument might  
>> be useful for those who dread a repeal, because if the method  
>> elects the CW, supporters of a single loser can't dress the  
>> complaint that the wrong candidate won up as a repeal of the  
>> method, simply because they don't have the voters required to make  
>> the repeal pass simply by that property alone. That is not what  
>> happened in Burlington, but it's similar - Condorcet minimizes this  
>> chance, and beatpath-based methods try to do so in the case of  
>> cycles as well.
>>
>> It should also ask the actual people, voters, what they think is  
>> important with respect to an election method, if such can be done.  
>> If simplicity matters, Ranked Pairs' relative simplicity may be  
>> more important than Schulze's track record, for instance. Asking in  
>> that manner could also help letting it know which arguments work -  
>> e.g. if the canceling-out phrasing of the singlepeakedness theorem  
>> gives a sense of fairness.
>
> as much as i like the Schulze method, since it is so much more  
> difficult to explain and for a lay person to comprehend, there will  
> always be some suspicion around it in the minds of people who want  
> to "Keep Voting Simple" (the signs from the IRV opponents in  
> Burlington).  they won't like any Condorcet, because most  
> fundamentally do not like the ranked ballot, but since Ranked Pairs  
> (which is simpler to understand) and Schulze pick the same winner if  
> there is a CW (the most common case, i believe) and if the Smith Set  
> is no larger than 3 (which, i believe will take up the other 1% of  
> cases), then i cannot imagine how it could be more efficacious to  
> promote Schulze Beatpath over Tideman Ranked Pairs.  but, also for  
> simplicity, maybe the best method to sell is simple Condorcet and  
> then, if no CW exists, pick the plurality (of first choices)  
> winner.  i'm not saying it's best, just that it's simple to  
> understand and that the likelihood that no CW exists will be small  
> (or is, at least, believed to be small).

If you went looking for a CW, seems like the list of cycle members  
should be easy to find.  These have identified themselves as more  
deserving than nonmembers.  Looking at A>B>C>A we know they have to be  
near in size - B large enough to support  B>C, yet small enough to  
support A>B.

I would stay away from first choices.  For one thing they are hard to  
identify if the race is multi-precinct.
>
> the other method, BTR-IRV (which i had never thought of before  
> before Jameson mentioned it and Kristofer first explained to me last  
> May), is a Condorcet-compliant IRV method.  i wonder how well or  
> poorly it would work if no CW exists.  i am intrigued by this method  
> since it could still be sold to the IRV crowd (as an IRV method) and  
> not suffer the manifold consequences that occur when IRV elects  
> someone else than the CW.  does "BTR" stand for "bottom two  
> runoff"?  and who first suggested this method?  is it published  
> anywhere?  Jameson first mentioned it here, AFAIK.  the advantage of  
> this method is that is really is no more complicated to explain than  
> IRV, and it *does* resolve directly to a winner whether a CW exists  
> or not.  i am curious in how, say with a Smith Set of 3, this method  
> would differ from RP or Schulze.

For Condorcet you have the N*N matrix and precinct summability but,  
unlike IRV, you better do nothing that involves going back to look at  
any ballots.
>
Dave Ketchum
>
> --
> r b-j                  rbj at audioimagination.com