[EM] publicly acceptable election methods

[Dave Ketchum]

Note that, while I formally respond to Russ, I also base this on James 
Green-Armytage words at Tue, 08 Mar 2005 02:31:46 -0500

On Mon, 07 Mar 2005 20:36:07 -0800 Russ Paielli wrote:

> Folks,
> 
> I realize that this email list is intended for a wide-ranging technical 
> discussion of election methods, and far be it from me to try to 
> discourage "brainstorming" for innovative new ideas.  Let me just 
> suggest, however, that it would be useful to distinguish between 
> election methods that could be publicly acceptable within our lifetimes 
> and those that almost certainly will not be.

So, here, we are considering what might be best of the practical 

possibilities for public elections.
> 
> I believe that Approval is simple enough to be publicly acceptable.


It may be that, but I choke on the deceptively simple to describe, yet 
tricky to use, decision as to what candidates to approve.

> IRV 
> may or may not be. 


I have to accept IRV as being simple for the voters for Condorcet, WHICH I 
DO WANT, uses the same ballot; and usually picks the same winner as IRV.

IRV has backers who sell it, loudly, as being beauty personified but:
      It has a spoiler problem, real though not as bad as Plurality's. 
With luck this will be exposed soon with voters KNOWING that A is most 
popular while B is less; that some A voters are ENTHUSED with C; and that 
IRV, stumbling over spoiler C, will declare B to be winner.
      IRV backers CLAIM it always finds a majority winner - misleading 
since readers are led to assume majority of the voters - which IRV fails 
to find in the above example - and it can happen that there was no such 
majority found - IRV can only claim a majority of the ballots remaining 
after exhausted ballots are discarded.
      Agreed something based on IRV is being accepted in VT, but what I 
found on their website had considerable tailoring.

> As I have pointed out before, IRV will have serious 
> implementation problems due to lack of "summability," but at least its 
> rules are relatively simple, with only minor variations, and no great 
> factions exist within its core advocates.
> 
> Unfortunately, the same cannot be said for Condorcet as it is currently 
> understood. Imagine how befuddled the public will be about "dropping 
> defeats" (as opposed to the much simpler idea of dropping candidates). 
> Do we use Schulze, Ranked Pairs, or some other variation? Then there's 
> the "margins vs. winning votes" issue, which is enough all by itself to 
> derail any potential progress once the intolerant dogmatists get 
> involved. (Will Mike declare a jihad on Condorcet if the public prefers 
> margins?)


Clearly we need to sort this into a single definition to sell:
      Probably makes sense to talk of a tournament, leaving the name 
Condorcet in the background.
      Usually the array identifies a single winner - takes a near tie of 3 
or more candidates to make a cycle.
      How resolve a cycle?  I say no complications for the voter, such as 
mixing in Approval.  Just analyze the array.
      We get excited over strategy - how often is this a problem in public 
elections, for which enemies likely find out and respond in kind?
      Margins vs WV?  I hear arguments for WV.
      Do we count =?  I say yes with 2 = giving each side credit for 1 
win.  Not greatly valuable except some voters will desire it.  I would 
credit the winner by 2 for > and <; each of the pair 1 for =.
      How resolve a true tie?  Customers likely want to define this for 
themselves, but I offer "flip a coin" to cover this unless they choose 
something.
      http://userfs.cec.wustl.edu/~rhl1/rbvote/desc.html - a reference of 
some possibilities.

> 
> I've also been reading a lot here about "lottery" methods. I have the 
> utmost respect for those who are discussing them, most of whom are 
> probably smarter than me, but I must confess that I don't have a clue 
> about what they are trying to accomplish. I think it is safe to say, 
> nevertheless, that any method that involves a "lottery" is publicly 
> unacceptable. We don't need a lottery now, so why should we adopt a 
> system that needs one? That's what the public will want to know.


Agreed the word "lottery" is unacceptable.  If the goal is actually 
needed, bury it in the program without talking about it.  Seems like a 
dozen candidates is about the limit for needing quick results.

> 
> If Condorcet is ever to be publicly acceptable, I think it needs to be 
> simplified. As most of you know, I recently proposed a 
> Condorcet-Approval hybrid method that has actually been suggested before 
> here on EM. It is much simpler than traditional Condorcet algorithms, 
> and I think it could possibly be effective *and* publicly acceptable.


Do not understand how mixing in Approval made this simpler for the voters 
(could be some convenience for program writers, but not believably enough 
to justify complicating the voters' tasks).

> 
> Let me briefly restate the idea:
> 
> The voters rank the candidates and also specify an Approval cutoff. The 
> CW wins if one exists, otherwise the least approved candidate is 
> repeatedly eliminated until a CW is obtained.
> 
> That's it, folks! Two sentences. Now, *that* could be simple enough to 
> be publicly acceptable. No complicated algorithms understandable only by 
> Ph.D.s are needed. If only the approved candidates are allowed to be 
> ranked, then it becomes even simpler to implement because then the voter 
> need not even specify a separate approval cutoff. If no equal rankings 
> are allowed -- simpler yet. These are all tradeoffs to be discussed, but 
> regardless of the final version it will still be much simpler than, say, 
> Schulze or Ranked Pairs.
> 
> I would really like to know what the exceptional folks here on EM think 
> about this proposal. Some of you have provided some feedback already, 
> and I certainly appreciate it. However, I don't sense any excitement 
> about it, and I'd like to know why not. What is the problem with it


Quoting James:  "2. It is not Smith-efficient; that is, it doesn't always 
choose a member of the minimal dominant set. I think that Condorcet 
methods that lack Smith efficiency are somewhat inelegant."

Not clear to me except, the voters can read the array - the declared 
winner BETTER be the CW or the winner of the cycle that the array identifies.

> 
> If this proposal is seriously flawed, then I seriously doubt that 
> Condorcet has a chance of ever being used in large-scale public 
> elections in my lifetime.


I claim that my words above make it more acceptable - just need to educate 
or steal the IRV sales crew.

> 
> --Russ

-- 
  davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.