[EM] Wikipedia

[Adam Tarr]

Tom Ruen wrote:

>Someone identified only by IP address modified the Wikipedia page:
>http://en.wikipedia.org/w/wiki.phtml?title=Special:Contributions&target=128.211.150.75
>
>Would someone like to take the credit?

I'll cop to it.  Although things had already been changed from what you had 
written when I got there.

>My category "one vote method" (where voters are limited to a single
>selection) was converted to a Up/Down vote method. It certainly looks
>strange to have 4 methods listed and 3 of 4 are "single votes" and 1/4 is
>"approval vote".

Which, of course, brings me back to my earlier complaint that "single vote" 
versus "multiple vote" is a false dichotomy.

I recently changed it to Yes/No, although both names seem equally 
reasonable to me.  The bottom line here is that if we are going to divide 
the methods by what ballots they use (a reasonable thing to do), then let's 
actually do that.  Every method that can be done by putting a check box 
next to each name is in the first category.

Similarly, every method that can be done by having as many slots as 
candidates is in the second category.  And every method that can be done by 
having some independent number of slots per candidates is in the third 
category.

>Approvally is undeniably a "Cardinal Rating subsystem", voters are given 2
>possible values and can assign them independently between all candidates.

It is also undeniably a yes/no system.  Every other rated voting system 
requires a different type of ballot from plurality.  Approval does not.

>I suppose the defense is that approval and plurality ballots look identical

Precisely.  And since the first line of the section was "Single Winner 
systems can be classified by ballot type" (which I did not write) it seems 
appropriate to actually classify things by ballot type.

>A subcategory "tied rankings permitted (or not)" was added to the Rank
>methods. I don't find that particularly useful since almost any methods may
>have considerations for ties and possibly different ways of counting them
>and I don't think there's a standard answer.

I'm not referring to ties in results there, but ties in rankings.  Which is 
a difference in how the voter votes, which is the overriding approach I was 
trying to use in classifying things.

I had earlier (in my 11:28 message) proposed "Top-preference prioritizing 
methods", "Condorcet compliant methods", and "other" as subdivisions in 
this area, but this really seems more arbitrary and less in keeping with 
the overall idea of classifying methods by ballot type.

>  I certainly don't think
>standard Bucklin allows tied rankings as categorizes below.

Bucklin has no difficulty handling tied rankings at all.  Surely some 
people describe it one way, and others describe it another way, yes?  As 
you said, it is fundamentally an approval-type method.

Is there some official version we should go by?  Bucklin isn't in use 
anywhere to my knowledge, so there's no real standard here.  If it is 
generally considered to not allow ties we can move it, but given that no 
additional rules are needed for it to handle ties, it seems kind of silly 
to do so.

>The issue of ties also brings out the difference between "single vote" and
>"multiple vote" systems in my original effort. IRV (and STV) can handle ties
>by "approval votes" OR "split votes", but only "split votes" will satisfy
>the "one vote, one person" ideal. You can't get Proportional Representation
>with STV by approval ties, but split-vote ties are valid, even if cumbersome
>to count.

The issues you are talking about here really only apply to 
multi-winner.  But it's worth noting that proportional approval voting gets 
proportionality out of "approval votes" through the use of a somewhat 
unique counting method.

>I'm done with my changes for now, but I respectfully ask others to evaluate
>these changes.

I agree that some feedback would be nice.  I didn't mean to ambush your 
changes.  I actually didn't think they were yours.  If others want to 
change my changes, that's fine too.

>1. Is "Up/Down" a good name for a ballot type? (I say it is just a hack to
>fit Approval into a category it didn't belong.)

It is a pretty lame name.  But it is definitely not a hack.  The Yes/No 
category is every voting system that can be carried out using the most 
common type of ballot - a list of names with one box next to each 
name.  How is that a "hack"?

>2. Does Approval belong more reasonably with "Ratings"? (I say no)

I agree, although I think you meant to say yes.  As I wrote in the wiki 
page, approval can be considered a rated method with only two choices.  It 
is the only method that can fit in two categories comfortably.  But since 
it works on the simplest sort of ballot, it seems unfair to put it in the 
category with range ballots.

>3. Do "Ties permitted" really deserve to be a subcategory?

Perhaps, perhaps not, but it was one of the few ways I could see to 
subdivide the many ranked methods.  And it is a distinction is based on the 
ballots used, which was the standard I was trying to keep.

>(I think a
>"counting type", distinguishing between single counts "Plurality, approcal,
>Borda" and multiple counts is more useful.)

And once again, I argue that this is an artificial distinction that is 
based on one approach for handling those ballots.  I can define a multiple 
count approach to handling approval ballots, which always produces the same 
winner yet has each voter casting at most one vote per round.  Does that 
mean that we should put that version (which has the same ballots, the same 
results, and the same academic criteria) in a different category from 
regular approval?

There may be a better distinguishing factor that allowing ties, but 
"counting type" is not it.

-Adam