[EM] Poll, preliminary ballots

[Richard, the VoteFair guy]

On 4/24/2024 1:46 PM, Chris Benham wrote:

 > On 25/04/2024 4:04 am, Richard, the VoteFair guy wrote:
 >> Here in Oregon we use a pen to fill in empty ovals on a paper ballot.

 > So why not change to writing numbers in boxes?  Aren't the
 > voters there numerate?

Numerate?  Probably some aren't.

Also remember some voters are reading the ballot in a non-English language!

Also remember that ALL ballots are counted by machine.  And fast.  There 
isn't time to deal with handwriting issues.

 > (And wouldn't that use less paper?)

Of course.  But asking voters to write numbers would be harder to "sell" 
to voters than "selling" ranked choice ballots (regardless of how they 
are counted).

I've heard Australians suggest boxes before.  THEY WILL NOT WORK HERE IN 
THE UNITED STATES!  For multiple reasons, including others I don't have 
time to explain.


 > ... that causes mono-raise (aka monotonicity) failure ...
 > ... meets Dominant Candidate ... and Later-no-Harm and
 > Later-no-Help and Clone-Loser and Mono-add-top and the Plurality
 > criterion ...
 > ... fails Clone-Winner and (therefore) Dominant Coalition ...
 > ... Condorcet criterion ...
 > ... fail Later-no-Harm, Later-no-Help and Mono-add-Top.

Pass versus fail criteria are meaningful to math savvy folks, but 
meaningless to voters.

What voters want is a method that rarely fails to elect who they think 
should have won.  Yes, that's subjective.

HOW OFTEN a method fails -- according to what they think is "common 
sense" -- is much more meaningful than pass-versus-fail!

I'll continue to repeat this point if you continue to talk about 
pass-fail criteria as if voters care about well-defined criteria.

This includes the Condorcet criterion, which not all voters think is the 
best kind of "majority" support the winner needs.  You can blame 
FairVote and Star fans for undermining the importance of the Condorcet 
criterion (in favor of other ways to judge "majority" support).

BTW, I used to believe the Condorcet winner always deserves to win.  And 
I reserve the right to claim that in the future.  But while I'm fighting 
misinformation from FairVote and Star fans I'm needing to relax the 
importance of Condorcet failures somewhat.


Again, thank you for writing clearly so that it takes me less time to 
answer.

Richard Fobes
The VoteFair guy


On 4/24/2024 1:46 PM, Chris Benham wrote:
> Richard,
> 
>> In Australia your voters write a ranking number (for each candidate) 
>> in a box.  So you only need one box per candidate.
>>
>> Here in Oregon we use a pen to fill in empty ovals on a paper ballot.
> 
> So why not change to writing numbers in boxes?  Aren't the voters there 
> numerate?  (And wouldn't that use less paper?)
> 
>> As a result, ballot "real estate" prevents us from printing as many 
>> choice columns as candidates. 
> 
> With that limitation my enthusiasm for both IRV and Benham evaporates. 
> Instead I would recommend Smith//Approval(implicit) with of course 
> equal-ranking allowed.  That might be my high Social Utility 
> bang-for-buck champion in any case.
> 
>> IRV's rule of assuming the candidate with the fewest highest-ranking 
>> marks is least popular is the "body" that isn't well designed.
> 
> Let me offer two versions of Hare (aka IRV) explaining its 
> motivation/justification as an improvement on FPP (aka the Single 
> Non-Transferable Vote, aka plurality).
> 
> *1. I see plurality has a "spoiler" problem and sometimes the winner 
> doesn't get a "majority". I'll fix that by one-at-a-time eliminating 
> weak candidates because I'm too stupid to realise that causes mono-raise 
> (aka monotonicity) failure, and my measure of weakness will be 
> favourite-of-the-fewest-among-remaining- candidates because I'm lazy and 
> simple-minded and too stupid to realise that causes failure of Condorcet.
> 
> So I really am a very badly designed and flawed method, so even STAR is 
> better and nearly every Condorcet method is much better in every way.
> 
> 2.  The SNTV meets Dominant Candidate (meaning that a candidate X that 
> is voted below no others on a number of ballots that is greater than X's 
> maximum pairwise opposition must win), and Later-no-Harm and 
> Later-no-Help and Clone-Loser and Mono-add-top and the Plurality 
> criterion.  In these ways I refuse to be worse. But unfortunately SNTV 
> fails Clone-Winner and (therefore) Dominant Coalition.  I will fix that 
> without losing any of those previously mentioned good properties.
> 
> So I am doing that in the only way possible. Therefore I am well 
> designed method and not "flawed".  Any modification of my algorithm or 
> any kludge stuck on me will cause me to lose some of my good properties. 
> For example forcing me to meet the Condorcet criterion will cause me to 
> fail Later-no-Harm, Later-no-Help and Mono-add-Top.  If you don't share 
> my criteria priorities you don't have to like me, but maybe you now have 
> some idea why some people like and/or respect me.*
> 
> Guess which one I think is correct.   "Dominant Candidate" and "Dominant 
> Coalition" are irrelevant ballot independent (stronger) versions of 
> Majority and Mutual Majority (aka Majority for Solid Coalitions) that I 
> coined.
> 
> Chris B.
> 
> 
> 
> On 25/04/2024 4:04 am, Richard, the VoteFair guy wrote:
>> On 4/22/2024 9:23 AM, Chris Benham wrote:
>>
>> > Your approach is like that of a quite bad and sloppy designer of a car
>> > or a plane. Every time it crashes you just stick another kludge on it
>> > designed to only guard against another crash just like the most 
>> recent one.
>>
>> Actually I'm moving a poorly designed car body (fenders, roof, doors, 
>> etc) from a poorly designed chassis (wheels, brakes, engine, drive 
>> train, etc) to a well-designed chassis (new wheels, new brakes, new 
>> engine, new drive train, etc).  Later we can replace the poorly 
>> designed body with a better-looking body.  Then we'll have a 
>> well-designed car.
>>
>> To clarify, eliminating pairwise losing candidates and using ranked 
>> choice ballots is the "chassis" in this analogy.  IRV's rule of 
>> assuming the candidate with the fewest highest-ranking marks is least 
>> popular is the "body" that isn't well designed.
>>
>>
>> > For IRV (aka Hare) I am strongly in favour of allowing unlimited strict
>> > ranking from the top. I was talking about a reasonable relatively 
>> benign
>> > way of dealing with equal-ranking in defiance of the ballot rules. In
>> > Australia I think the whole ballot is not counted, and binned as
>> > "informal".  Normally there should be nothing stopping you from ranking
>> > the most evil candidates strictly below all the others.
>>
>> In Australia your voters write a ranking number (for each candidate) 
>> in a box.  So you only need one box per candidate.
>>
>> Here in Oregon we use a pen to fill in empty ovals on a paper ballot.
>>
>> Too many ovals (beyond about seven) per candidate inflates the paper 
>> ballot size to unreasonable dimensions.  Already, with just one oval 
>> per candidate, the ballot covers both sides of a large paper ballot, 
>> and sometimes there are two ballot pages.
>>
>> As a result, ballot "real estate" prevents us from printing as many 
>> choice columns as candidates.
>>
>> This limitation, plus the silly rule of not correctly counting two or 
>> more marks in the same choice column -- so-called "overvotes" -- stops 
>> us from being able to rank all other candidates above our 
>> most-disliked candidate.
>>
>>
>> > Assuming you can pair up all the "over-voting" ballots in this way, 
>> this
>> > seems to be equivalent to dividing the votes up into equal fractions
>> > that sum to 1.
>>
>> Yes, those of us who understand math recognize that decimal numbers 
>> work fine.  But few voters, and very few politicians, understand math. 
>> Especially fractions and decimal numbers.
>>
>> > But what if you can't pair them all off, or someone votes
>> > more than two candidates at the same ranking level?
>>
>> That error is almost similar to truncating the decimal numbers to the 
>> nearest smallest integer.  The "almost" refers to a few ballots that 
>> can't be "paired up with" an equivalent preference pattern.
>>
>> That "pairing" also works with three ballots with the same three 
>> top-ranked candidates.  And it works with four ballots ranking the 
>> same four candidates highest.  Etc.
>>
>>
>> >>> If there are "failure types" in need of names, what's stopping you
>> >>> from giving them names? ...
>>
>> >> Time and money.
>>
>> > I wasn't talking about for the purpose of discussions in the
>> > mass media or to get text books or dictionaries changed.  I was
>> > just talking about just for the purpose of (hopefully
>> > somewhat rigorous) discussion here.
>>
>> My time is still a huge limiting factor.  I'm juggling lots of 
>> projects.  That's why I don't have time to reply to as many messages 
>> here as I'd like.
>>
>>
>> Chris, I'm grateful that your messages are well-written.  That makes 
>> them easier to reply to.  Thank you for taking the time to write clearly!
>>
>> Richard Fobes
>> The VoteFair guy
>>
>>
>> On 4/22/2024 9:23 AM, Chris Benham wrote:
>>> Richard,
>>>
>>>> Horrible, yes.  Garbage, no, because STAR a clever way to improve
>>>> single-winner score voting.
>>> It trashes Score voting's compliance with Favorite Betrayal and 
>>> Participation to gain merely Condorcet Loser. Pure genius. If it is 
>>> an attempt to "improve" Score voting (which I have great difficulty 
>>> believing) then I don't agree that it qualifies as "clever".
>>>
>>>> > If there are "failure types" in need of names, what's stopping you 
>>>> from
>>>>   > giving them names? ...
>>>>
>>>> Time and money.
>>> I wasn't talking about for the purpose of discussions in the mass 
>>> media or to get text books or dictionaries changed.  I was just 
>>> talking about just for the purpose of (hopefully somewhat rigorous) 
>>> discussion here.
>>>
>>>> Our goal is to rise way above plurality.  Accepting limitations of
>>>> plurality is unnecessary.
>>>>
>>>> Why impose any extra strategic burden on the voter?
>>>
>>> I agree that that should be avoided. As you would know if you read my 
>>> previous posts here about STAR, the strategic burden it places on the 
>>> voter is vastly greater than the one imposed by plurality (aka FPP).
>>>
>>> Both have Compromise incentive while STAR also has very strong 
>>> Push-over incentive.
>>>
>>>> > I think it is reasonable to treat ballots that (against the ballot
>>>>   > rules) equal-rank above bottom as though they truncated at that 
>>>> point.
>>>>   > Doing otherwise (as I think you advocate) without a quite complex
>>>>   > procedure I suggest makes the method a bit more vulnerable to 
>>>> Push-over
>>>>   > strategy.
>>>>
>>>> I and most voters want to be able to rank an evil candidate -- Gollum,
>>>> Voldemoron, etc. -- below all other candidates.  Truncation means the
>>>> evil candidate is as acceptable as other "bad" candidates.
>>>>
>>> For IRV (aka Hare) I am strongly in favour of allowing unlimited 
>>> strict ranking from the top. I was talking about a reasonable 
>>> relatively benign way of dealing with equal-ranking in defiance of 
>>> the ballot rules. In Australia I think the whole ballot is not 
>>> counted, and binned as "informal".  Normally there should be nothing 
>>> stopping you from ranking the most evil candidates strictly below all 
>>> the others.
>>>
>>>> I've written code that correctly counts so-called "overvotes."  It's 
>>>> not
>>>> a "complex procedure":
>>>>
>>>> https://github.com/cpsolver/VoteFair-ranking-cpp/blob/master/rcipe_stv.cpp
>>>>  When shared preference levels are encountered,
>>>> //  the ballots are transfered in "whole" numbers,
>>>> //  not by splitting a ballot into fractional or
>>>> //  decimal portions.  For example, during a
>>>> //  counting cycle, if there are two ballots that
>>>> //  rank candidates numbered 1 and 2 at the same
>>>> //  highest ranking level, one of the ballots will
>>>> //  transfer to candidate 1 and the other ballot
>>>> //  will transfer to candidate 2.
>>>> //
>>>
>>> Assuming you can pair up all the "over-voting" ballots in this way, 
>>> this seems to be equivalent to dividing the votes up into equal 
>>> fractions that sum to 1. But what if you can't pair them all off, or 
>>> someone votes more than two candidates at the same ranking level?
>>>
>>> I didn't express myself quite clearly enough. The "complex procedure" 
>>> I referred is the one I, not you, suggest.  I didn't bother 
>>> describing it.
>>>
>>> I think that if we allow above-bottom equal-ranking in IRV or Benham, 
>>> then if among remaining candidates some ballots rank more than one 
>>> candidate equal-top then we make a provisional order of the 
>>> candidates by counting those ballots as equal fractions summing to 1.
>>> (A=B counts as half a vote to each of A and B, A=B=C counts as a 
>>> third of a vote to each of A and B and C, and so on.  Now it would be 
>>> fine for this to be the final order for deciding which candidate to 
>>> next eliminate were it not for the fact that it makes Push-over 
>>> strategising easier.)  Then we count the equal top (among remaining 
>>> candidates) ballots again, this time they give a whole vote to 
>>> whichever of the ones they equal rank to the one that was highest in 
>>> the provisional order. (So an A=B ballot gives a whole vote to 
>>> whichever of A and B was higher in the provisional order, and of 
>>> course nothing to B.)
>>>
>>> This is fully in the spirit of the Single Transferable Vote but I 
>>> think you will agree that it is complex. I don't think allowing 
>>> above-bottom equal-ranking in those methods is so important, nor do I 
>>> think there would be any significant demand for that from voters, so 
>>> I don't advocate allowing it for those methods.
>>>
>>>> I'm bothered by the failures in Burlington and Alaska.  But those were
>>>> not just Condorcet failures.  They also were IIA failures,
>>>> center-squeeze failures, etc.
>>> It is the most basic theory that all remotely reasonable methods 
>>> fail  IIA, so why are we even mentioning that?  And isn't 
>>> "center-squeeze" just a vague concept used in anti-IRV propaganda? 
>>> What is the precise definition of a "center-squeeze failure"?
>>>
>>> Your approach is like that of a quite bad and sloppy designer of a 
>>> car or a plane. Every time it crashes you just stick another kludge 
>>> on it designed to only guard against another crash just like the most 
>>> recent one.
>>>
>>> Chris B.
>>>
>>> *Richard, the VoteFair guy*electionmethods at votefair.org 
>>> <mailto:election-methods%40lists.electorama.com?Subject=Re%3A%20%5BEM%5D%20Poll%2C%20preliminary%20ballots&In-Reply-To=%3C632ea079-e977-441c-bf19-41522d2d8eee%40votefair.org%3E>
>>> /Sat Apr 20 10:30:57 PDT 2024/
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> On 4/19/2024 1:15 AM, Chris Benham wrote:
>>>
>>>   > ... It is not garbage like STAR.
>>>   > ...
>>>   > ... STAR is a horrible method that is very highly
>>>   > vulnerable to both Compromise and Pushover.
>>>
>>> Horrible, yes.  Garbage, no, because it's a clever way to improve
>>> single-winner score voting.  It's useful among friends when voting is
>>> not anonymous.  Or when "dishonest" exaggeration cannot be hidden.
>>>
>>>   > If there are "failure types" in need of names, what's stopping 
>>> you from
>>>   > giving them names? ...
>>>
>>> Time and money.  Unlike two STAR promoters, the folks at FairVote, and
>>> academic professors, I'm not getting paid to promote or advance
>>> election-method reform.
>>>
>>>   >>   Approval voting requires tactical voting.  There's no way to 
>>> avoid it.
>>>   > The strategic burden on the voter is certainly no greater than with
>>>   > FPP.  ...
>>>
>>> Our goal is to rise way above plurality.  Accepting limitations of
>>> plurality is unnecessary.
>>>
>>> Why impose any extra strategic burden on the voter?
>>>
>>>   >> Another difference from IRV is about what FairVote calls 
>>> "overvotes."
>>>   >> RCIPE counts them correctly.  ...
>>>   > I think it is reasonable to treat ballots that (against the ballot
>>>   > rules) equal-rank above bottom as though they truncated at that 
>>> point.
>>>   > Doing otherwise (as I think you advocate) without a quite complex
>>>   > procedure I suggest makes the method a bit more vulnerable to 
>>> Push-over
>>>   > strategy.
>>>
>>> I and most voters want to be able to rank an evil candidate -- Gollum,
>>> Voldemoron, etc. -- below all other candidates.  Truncation means the
>>> evil candidate is as acceptable as other "bad" candidates.
>>>
>>> I've written code that correctly counts so-called "overvotes." It's not
>>> a "complex procedure":
>>>
>>> https://github.com/cpsolver/VoteFair-ranking-cpp/blob/master/rcipe_stv.cpp
>>>
>>>   >> Avoiding any failures in REAL elections is what I'm "buying" by
>>>   >> advocating RCIPE instead of IRV.
>>>   > I'm still baffled as to why, if you don't like Condorcet 
>>> failures, you
>>>   > don't simply advocate a Condorcet method. How is the argument "Let's
>>>   > lose strict compliance with several criteria met by IRV so that 
>>> we can
>>>   > somewhat more often elect the Condorcet winner" better than
>>>   > "Let's lose strict compliance with several criteria met by IRV so 
>>> that
>>>   > we can ALWAYS elect the Condorcet winner"??
>>>
>>> I'm bothered by the failures in Burlington and Alaska.  But those were
>>> not just Condorcet failures.  They also were IIA failures,
>>> center-squeeze failures, etc.
>>>
>>> I want fewer failures in real elections.  I don't care about convoluted
>>> scenarios that would never occur in a real election.
>>>
>>> Again, thank you for this useful discussion.  I appreciate that you
>>> really want to understand why I rank some methods better than others.
>>>
>>> Richard Fobes
>>> The VoteFair guy
>>>
>>>
>>> ----
>>> Election-Methods mailing list - see https://electorama.com/em for 
>>> list info
>> ----
>> Election-Methods mailing list - see https://electorama.com/em for list 
>> info