Why is IRV being considered seriously here?

[Stephen Unger]

I find it strange that there is so much discussion here in which IRV
(Instant Runoff Voting) is treated as tho it deserved serious
consideration as a voting system. There is much debate about the
extent to which it favors centrists or extremists, or the
circumstances under which it encourages strategic voting. It appears
that some in this group may actually believe that there are
circumstances under which IRV might be a reasonable choice. All this
despite the fact that members of this group should be familiar with
the various ways in which IRV can exhibit bizarre behavior (with
non-miniscule probabilities).

Have people forgotten about the following IRV characteristics?

1. If all votes are reversed, the same candidate might still win. More
generally, reversing every vote will very often not lead to the order
of the results being reversed.

2. Possible failure to elect a candidate that IRV elections would show
as beating every other candidate in a 2-candidate election.

3. The possibility that top-ranking a candidate can cause that
candidate to lose.

4. Ranking a candidate last rather than first can lead to that
candidate winning.

5. That supporters of a middle-ranked candidate might have an
incentive to downgrade their favorite to prevent the election of a
candidate they feel is terrible.

A root cause of these problems is that, in processing the votes, IRV
considers only a portion of the voters' inputs. E.g., the second-place
choices of most voters are often ignored.

Nothing like this exists for score or approval voting, where all voter
inputs are always used to generate outcomes. At worst, there might be
cases in which voters may have to agonize as to the how much support
to give candidates they consider intermediate.

Furthermore, along with other systems based on ranking, IRV does not
allow voters to distinguish between

(1) Considering A as vastly better than C and the even worse D,

and

(2) Considering A excellent, a little better than B, with C considered
horrible.

This leads to cases such as A>B>C>A.

Furthermore, there is the major point that tabulating the results of
an IRV election cannot be done in precincts: it must be done
centrally, or there must be 2-way communications between precincts and
a processing center. This leads to great vulnerability to fraud both
in the communications process, and in the fact that there will be
enormous pressure to use computers in order to cope with the
complexity of the process.

I understand that IRV is brilliantly championed by a politically savvy
group, and is indeed the most widely used alternative voting
system. When consider superficially it looks very reasonable. This is
why so many good people favor it. I am sure that the Fairvote folks
started backing IRV with the best intentions. Sadly, they have not yet
found the strength to admit they made an honest mistake and to change
course. This makes it particularly important for us to publicize as
widely as possible the fundamental irrationality of IRV.

Steve
............

Stephen H. Unger
Professor Emeritus
Computer Science and Electrical Engineering
Columbia University
............

[Warren D Smith]

Not a bad summary. And IRV has even more flaws than Unger mentioned.
Its observed behavior of (apparently always) soon yielding massive
2-party domination in every country it is used in (within the body
elected by IRV) is devastating.
(E.g Australia, last 4 house elections cycles in a row, total number
of third party candidates elected: 1 in 600 races.)

The fact the voters in the top IRV-using country Australia say by
large margin they'd prefer to get rid of it, while also the voters in
the largest country ever to consider switching to IRV (the United
Kingdom) rejected it by landslide, also convincingly
demonstrates that IRV does not even have
political/propaganda/enactability advantages:
http://www.rangevoting.org/WhatVotersWant.html
IRV also is more complicated and expensive than rival systems for both
the counters and the voters.

The recent imbroglio with Rob Richie, the USA's foremost IRV
propagandist, attempting
to criticize Approval Voting, dramatized how little he has going for
him, intellectually
(and also how little he cares about the truth):
http://rangevoting.org/RichieOnApproval.html

[Clay Shentrup]

Steve,

I'd ask you to stop wasting time talking about election method criteria, and focus on Bayesian regret. IRV does pretty poorly compared to other alternative systems, but is markedly better than Plurality Voting, and comparable with delayed runoff. It's a pretty terrible system, compared to what we could easily have with Approval Voting or Score Voting. I'm not particularly concerned with it though.

I'm more concerned with the horrendously fallacious propaganda coming from IRV's major advocacy organization, FairVote. The damage they've done to democracy is incalculable. But sans FairVote, IRV itself is not a big deal. The biggest impediment to Approval Voting and other superior systems is general apathy and lack of knowledge about the very existence of "alternative voting systems".

[Warren D Smith]

I don't think election method criteria are a "waste of time." I agree
Regret is
foundationally superior -- but criteria still have their points, such
as increasing
human understanding and not requiring a computer.



--
Warren D. Smith
http://RangeVoting.org <-- add your endorsement (by clicking
"endorse" as 1st step)

[Stephen Unger]

Bayesian Regret (BR) is very valuable in evaluating election
systems. But, in some cases, it is a form of overkill.

Consider a system that has excellent properties with respect to
dealing with the spoiler problem, and various issues concerning
strategic voting. However, suppose the system is "noisy", in that for
about 25% of elections, it generates random results. In such a case, I
think it would be a waste of time to consider this system, and arguing
about whether or not it really has the good properties would be
pointless. There would be no need to bring in BR, which requires a lot
of explanation, and discussion of simulations. IRV fits this
description (actually, even without the noise, it is not as good as
the system described here.)

I don't see how you can brush off IRV as "not a big deal". Many good
people, who should know better, consider IRV to be a valuable
reform. We need to educate them about its weird behavior.

A few months ago I gave a talk to Columbia University Engineering
School freshman about voting systems and e-voting. One of my examples
was a simple case where IRV found the best candidate to be the same as
the worst candidate. This amazing behavior produced a pronounced,
strong reaction from the class, and I'm sure that they will all
remember this property of IRV.

Steve
............

Stephen H. Unger
Professor Emeritus
Computer Science and Electrical Engineering
Columbia University
............

> --
> You received this message because you are subscribed to the Google Groups
> "The Center for Election Science" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to electionscien...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
> �
> �
>
>

[George Sanders]

My gold standard for evaluating ANY voting method: 2 polarizing demagogues vs. everyone's favorite 2^nd choice. Common sense tells us we'd rather have 80% satisfaction ALL THE TIME rather than 100% satisfaction but only for HALF THE TIME. (Talk about your Bayesian Regret!) FPTP, IRV, and all the ranked voting methods fail this standard test. Only the rated voting methods (as far as I know) can provide us with the compromise candidate that would provide the greatest net satisfaction in this extreme case. But I think going to this extreme [figuratively] proves the social choice. If my common-sense logic here meets with your approval [intended], then perhaps this may be the best way to explain to the ‘common-man’ electorate at large just why Score Voting (and its easiest-to-implement, simplified cousin—Approval) are so vital in ameliorating partisan politics: it changes the ‘majority wins’ concept of It’s US against THEM into a communal spirit of We’re ALL in this TOGETHER. So I ask you then: considering the world situation, could there be anything more important for World Peace other than the way in which We, the People, allow ourselves to be governed?

P.S.: Roseanne Barr favorited my recent Tweet:
@TheRealRoseanne  Thanks for your prior re-Tweets. I now have a 3-minute video: best simple way to end partisanship! https://www.youtube.com/watch?v=V4LqypgNIOw …

> �
> �
>
>

[Michael Ossipoff]

 

 

Why is IRV being seriously considered here?

 

Considered for what? There are 3 different positions on IRV that have been expressed here:

 

1. All-ouit opposition to IRV, without regard to conditioins.

 

2. FairVote representatives have spoken here too.

 

3. My claim that IRV and Approval each better than eachother, for different conditions.

 

It would be difficult or impossible to answer Steve's question without more detailed specification of what consideration he's referring to.

 

 

On Saturday, May 18, 2013 12:12:01 PM UTC-4, Steve wrote:

 

 

I find it strange that there is so much discussion here in which IRV
(Instant Runoff Voting) is treated as tho it deserved serious
consideration as a voting system.

 

 

Steve is surely referring to the dedicatedly anti-IRV people who were telling the reasons why they don't like IRV.

 

Funny, I didn't hear them considering IRV.What were they considering proposing it for?

 

But, to use "consider" a little differently, surely you understand that your criticism of something won't count for much if it's clear to people that you dont, give it fair and serious cnsideration. If you don't, then you're just engaging in knee-jerk assertion and monologue.

 

 

There is much debate about the
extent to which it favors centrists or extremists

 

That's an over-dramatic way of saying that it can fail to elect a CW, as can Approval. IRV can do so automatically, when no one intended to leave that CW unprotected, and so it can fairly be said that Approval is better-disposed to elect CWs than IRV is.

 

When IRV elliminates a CW, it elects someone who receives the CW's transfers. It elects someone who is the next choice of the CW-preferrers. You conveniently forget that when you hysterically warn that iRV elects extremists.

 

 

, or the
circumstances under which it encourages strategic voting.

 

That would refer to IRV's FBC failure. Under current conditions that's a prohibitive problem, a reason why IRV is anadequate for current conditions.

 

IRV advocatres and Approval/Score advocates remind me of the two Liliputian nations that were warring about the issue of which end of a hard-boiled egg should be opened first.

 

I've tried to explain this, but neither side seems to get it.

 

Under current conditions, with disinformational media, and a public who believe whateverr they say, IRV is entirely inadequate, and Approval or Score would be much better than IRV.

 

But with honest, open, participatory and agenda-free media, and an electorate who were no longer complete suckers, FBC wouldn't be necessary. Then, if you're concerned about strategy, IRV would be better than Approval. Yes, that's forbidden to say here.

 

IRV, but not Approval, meets MMC, and is free of the chicken dilemma. Neither of those things can be said for Approval or Score.

 

IRV, but not Approval meets Tideman's Clone Independence Criterion.

 

In other words, both sides of the Liliputionan war are quite wrong. The Approvalists and the IRVists are wrong.

 

Maybe I haven't emphasized well enough before now.

 

 

It appears
that some in this group may actually believe that there are
circumstances under which IRV might be a reasonable choice.

 

 

Funny, I haven't heard anyone here but me say that. Of course my name is unmentionable here :-)   To suggest that Approval and Score aren't always the best is contrary to the party-line here.

 

Wrong. In the Green scenario, IRV isn't just a reasonable choice. In the Green scenario, IRV is an excellent choice.

 

(even though Benham or (especially) Woodall might be more practical because of not sharing IRV's vulnerability to replacement due to IRV's ability to fail to elect a CW)

 

 

All this
despite the fact that members of this group should be familiar with
the various ways in which IRV can exhibit bizarre behavior (with
non-miniscule probabilities).

 

Members of this group should be aware that every voting system fails criteria. In fact every voting system fails criteria that are obviously desirable.

 

You give up one criterion in order to gain another one.  Members of this group should understand that.

 

Have people forgotten about the following IRV characteristics?

1. If all votes are reversed, the same candidate might still win. More
generally, reversing every vote will very often not lead to the order
of the results being reversed.

 

No one has answered my question about why that's important. How does it cause  strategy problem that could result in societal harm?

 

I'm going to give this answer to all of the criterion-failures that Steve trots out in this post:

 

IRV meets MMC and has no chicken dilemma. That's an extremely powerful combination of properties. (What's that? Have I already said that?)

 

It means that voters in a mutual majority need only rank sincerely, to guarantee that any mutual majority to which they belong will automatically, without any strategy, elect someone from their MM-preferred set. Steve is  conveniently missing that. That's the rank-balloting ideal, and IRV makes it available to members of a mutual majority.

 

If you want to enforce _all_ majorities, however constituted, then choose a CC-complying method, such as Benham or Woodall, or Schwartz Woodall.

 

Obviously every desirable property or combination of properties comes at a price. So what? You already knew that.

 

For current conditions, choose Approval, Score, Symmetrical ICT, or ICT.

 

For the Green scenario, choose IRV, Benham, Woodall, or Schwartz Woodall. Or, if you don't want to include non-MM voters in the choice among the MM-preferred set, then there's MM-Benham and MM-Woodall.

 

2. Possible failure to elect a candidate that IRV elections would show
as beating every other candidate in a 2-candidate election.

 

Yes, IRV fails the Condorcet Criterion (CC).  Approval fails CC too. So does Score.

 

 

3. The possibility that top-ranking a candidate can cause that
candidate to lose.

 

IRV and all Condorcet methods fail Mono-Add-Top, and Mono-Add-Unique-Top.  So what?

 

Sure, I admit that I've used those criteria to criticize Condorcet methods, when arguing for Approval for Score under current conditions. But the monotonicity criteria, such as Mono-Add-Top, Mono-Add-Unique-Top, and Mono-Raise don't measure for strategy problems.

 

I'm not saying that Approval/Score has an insuperable strategy problem. But of course it has one. When FBC isn't needed, and when voters are concerned about strategy, then choose a rank method that's more strategy-free.   iRV is one of those.

 

Remember that IRV's only strategy problem is its FBC failure, something that doesn't matter under Green scenario conditions. (But, as I said, its CC-failure could make it vulnerable to replacement,  regardless of how good it is).

 

 

4. Ranking a candidate last rather than first can lead to that
candidate winning.

 

Mono-Raise failure. It doesn't measure for a strategy problem. Strategy problems are what matter most, societally.

 

I'm not saying that social utity doesn't matter at all. It's a reason why Woodall is better than Benham.

 

 

5. That supporters of a middle-ranked candidate might have an
incentive to downgrade their favorite to prevent the election of a
candidate they feel is  terrible.

 

FBC failure. Intolerable under current conditions. Not a problem under Green scenario conditions.

 

A root cause of these problems is that, in processing the votes, IRV
considers only a portion of the voters' inputs. E.g., the second-place
choices of most voters are often ignored.

 

Yes, that's a root cause of IRV's important freedom from chicken dilemma, resulting from its LNHa compliance.

 

Nothing like this exists for score or approval voting, where all voter
inputs are always used to generate outcomes.

 

Steve is falling into the trap of using a method's rule as a standard by means of which to support that method.

 

 

Lilliputians! You don't have to fight!

 

 

At worst, there might be
cases in which voters may have to agonize as to the how much support
to give candidates they consider intermediate.

 

Oh, "agonize", is that all!  :-)

 

Steve has said it well.  Without current-conditions' FBC need, and when strategy is felt to be important, Approval strategy could call for some agonizing. Not so IRV, if you're in a mutual majority.

 

And if you don't think your platform will be in a mutual majority's MM-preferred set, then maybe you should reconsider your platform. Sorry, but it isnt my fault, or IRV's fault, if you don't have a mutual majority who like your proposals. 

 

Furthermore, along with other systems based on ranking, IRV does not
allow voters to distinguish between

(1) Considering A as vastly better than C and the even worse D,

and

(2) Considering A excellent, a little better than B, with C considered
horrible.

 

First Steve complains that IRV (like Approval) fails CC. Then he cites as a disadvantage the count that makes it possible to meet CC.

 

Steve is implying preference for sincere rating in Score. in amicable organizations, voting on matters whose alternatives are quite similar, with none really odious, sometimes Score or Approval could be the best choice. In official public elections, in current conditions, Approval's and Score's FBC compliance is needed.

 

This leads to cases such as A>B>C>A.

 

That can happen in all Condorcet methods, and it has no effect on Woodall's and Benham's MMC compliance, freeom from chicken dilemma, and unmatched freedom from strategy-need for MM-members.

 

A top-cycle can spoil MMC compliance in Symmetrical ICT, and that is a disadvantage. I don't propose Symmetrical ICT for the Green scenario. 

 

 

Furthermore, there is the major point that tabulating the results of
an IRV election cannot be done in precincts:

 

Of course that's true of all methods, including Approval, Score, and Plurality. In all voting systems, it's necessary to do a central tabulation. Anti-IRVists conveniently forget that.

 

 

it must be done
centrally

 

 

...as is the case with Approval and Score as well.

 

 

 

, or there must be 2-way communications between precincts and
a processing center.

 

Of course. There always is 2-way communication with the processing center in our Plurality elections too. There would be in Approval as well.

 

An IRV count, as regards communication with the processing center would amount to nothing other than up to N Plurality elections, where there are N candidates.

 

 

This leads to great vulnerability to fraud both
in the communications process

 

Steve's right, if that communicatioin is done via secret couriers and private phone-lines  :-)

 

But howabout by broadcast and posted count results, just as we already always have with Plurality.

 

 

, and in the fact that there will be
enormous pressure to use computers in order to cope with the
complexity of the process.

 

Of course any count would be easier by computer. IRV could use it more than Approval. Woodall could use it more than IRV. So what?

 

I've ampy discussed how computer count fraud could be made entirely impossible.

 

A hint: Public imaging.

 

Look, if you just want to say that Approval and Score are the best proposals for current conditions, and that IRV is inadequate for current conditions, then just say that.

 

But you IRV and Approval/Score Liliputions needn't keep waging your war of hyperbole against eachother.

 

Michael Ossipoff

 

 

 

[George Sanders]

My problem with IRV: It's just another "majority wins" voting method (where a small majority can tyrannize a large minority). So, what's so special about going from 49.99999% to 50.00001%--did something 'magic' just happen?? We who propose Approval over IRV are proposing a voting method for the common-sense compromise candidate over any polarizing "majority winner" that may be forced up from the alternate choices of the so-called 'worst' candidate (the one who got the least first-place votes). And let's not forget: Approval Voting (imperfect a method as it is)--while being the easiest/cheapest/fastest way to begin to mitigate partisan politics--is just the stepping stone to Score Voting. I am approaching 70, but every time I explain Score Voting to the younger generation, with its parallel to the ratings on internet reviews, they get it INTUITIVELY and IMMEDIATELY!  And hey--don't our internet search-engine algorithms make the most use of Score Voting rather than any other method to project the selections likely to provide the users [voters] with the greatest net satisfaction?         

[Michael Ossipoff]

On Saturday, May 18, 2013 10:08:57 PM UTC-4, George Sanders wrote:

 

My problem with IRV: It's just another "majority wins" voting method (where a small majority can tyrannize a large minority).

 

Automatic majoriity rule is IRV's (and Woodall's and Benham's) big advantage. Why do we want that? Whether we like it or not, there are voting situations (such as all of our official pulbic elections) in which stratgey matters. We want the best that we can get. In Approval or Score, we must try to guess who's winnable, and what we can gett. Deal with the chicken dilemma. Support all of the acceptables, and somehow later judge, from vote totals, if we still need our compromise. Or try strategic fractional rating. It can be done, but it will take a few Approval elections to find out how well we can do, and to eventually get the best that we can. The beauty of automoatic majorty rule is that it's available in the first election. Automatic majoriy rule immediately, automatically, gives us what we'd eventually gradually accomplish strategically with Approval or Score.

 

But I don't advocate IRV, Benham or Woodall for current conditions. For current conditions, Approval and Score are the best poposals.

 

I don't think reform will happen at all under current conditions. I suggest that IRV will be the next voting system, because that's what the Greens offer in their platform.

 

 

 

 

 

So, what's so special about going from 49.99999% to 50.00001%--did something 'magic' just happen??

 

We'll be trying for that same majoriity rule with Approval or Score. The only difference is that it will take longer.

 

 

We who propose Approval over IRV are proposing a voting method for the common-sense compromise candidate over any polarizing "majority winner"

 

The CW is the natural comromise. Benham and Woodall always elect the CW when there is one. IRV doesn't, but if a majority don't like that (and there will be a dis-satisfied majority whenever IRV fails to elect the CW), then they'll, at that time, replace IRV with Benham or Woodall. So what's the problem?

 

It's essential to distinguish current conditions from the better conditions that I call "the Green scenario". For current conditions, where FBC is needed, and simplicity would bring better enactability (as if any reform were enactable under the Republocrats), Approval and Score are the best proposals for current conditions.

 

George is quite right about this: When IRV eliminates a CW, then, though IRV still chooses from the MM-preferred set, IRV's majority rule is, in that instance, only granted to MM members, and denied to other majorities, differently-constituted. That's uncompromising and inimical. IRV may very well be replaced by a dis-satisfied majority, for that reason. That happened in Burlington, though they replaced IRV with somethinig worse. In the Green scenario, if IRV is replaced, it will be with something more deluxe, more compromising--Benham or Woodall. They assure majority rule to all majoriities, however constituted.

 

Michael Ossipoff

 

[Abd ul-Rahman Lomax]

At 11:54 AM 5/18/2013, Warren D Smith wrote:
>The recent imbroglio with Rob Richie, the USA's foremost IRV
>propagandist, attempting
>to criticize Approval Voting, dramatized how little he has going for
>him, intellectually
>(and also how little he cares about the truth):
> http://rangevoting.org/RichieOnApproval.html

Warren Smith generally has sound opinions on voting systems, but not
about people.

Richie made, in an ad-hoc comment on our approval video, an error in
reading sources, causing him to misunderstand the total number of
voters in a University of Colorado Student government. He also made
in that section, a minor arithmetic error (of little significanc).
The "total voters" issue is not actually yet resolved, there have
been incorrect assumptions made by *everyone* who has discussed this.
We don't even know, and it's unlikely, that the election in question
was an Approval election, one interpretatation of the data that fits
the fact is that it was not, and that therefore, with an on-line
voting system, the sum of votes for two candidates was equal to the
number of voters. But there are other possibilities.

That video comment was civil, conciliatory, and can be read as a
reaching toward cooperation. Clever plot: pretend to be attempting
cooperation, and then your enemy will look really bad if he calls you a liar.

I prefer to read Richie at face value. He's fairly understandable. He
is a political activist, not a mathematician or, really, a general
voting systems expert (though he knows a lot, it's also been filtered
through his agenda, lots of people do that, especially activists). He
was, at one time, haughtily dismissive of election science. Perhaps
he's older and wiser now. Let's meet him in the future, rather than
in the past.

IRV is being seriously considered here because it is on the table for
the public, still. We know the problems of IRV. We know there are
more effective ways of accomplishing the goals of IRV. We haven't
forgotten that. Still, IRV uses the STV method of amalgamation, which
makes much more sense in a multiwinner environment. Again, most of us
think there are better ways, but there are forms of STV that many of
us could support.

Why would we support something that wasn't optimal? Well, the same
reason most of us support Approval. It's a possibility, and it's an
improvement over Plurality. (Additionally, with Approval, it's
termninally simple.)

I would never suggest replacing IRV with Approval, like some here
suggest. I'd suggest replacing IRV with Bucklin. Same ballot, and an
amalgamation method without the serious problems of IRV. Richie, in
his video comment, actually raises the best argument for IRV, lack of
LNH failure. I would -- and will -- argue that LNH failure is
actually a positive trait, the way that the "fear of harming their
favorite" causes Approval voting, and especially the more-flexible
Bucklin Voting, to more accurately measure preference strength than
one might think.

Basically, a fear of harming the favorite must be based, to be
controlling, on high preference strength. And that will cause a voter
to push a second preference down the ranks. Which is precisely an
expression of utility to the voter. It makes the system more accurate.

A *huge mistake* that many of us made was to consider Bucklin a
ranked voting system, and to analyze it as one. It's a Range system,
actually, as Jameson Quinn has well recognized, it's median range,
using only the approved range. It never formally shows the range
ratings as such, but that could easily be done, and the range could
be extended to full range, for fuller analysis.

Plurality vs. Approval. Same ballot, more flexibility in voting.
IRV vs. Bucklin. Same ballot, instant runoff approval. IRV:
eliminations, Bucklin, no eliminations, all votes are counted to the
rank where a majority of approvals is reached. (I would strongly
argue that all votes should be reported, even if not necessary for
determining the winner. It's rude to ask people to vote and not to
count the votes. But those lower ranked (rated) votes will provide
valuable information to candidates about their standing with the voters.

[Abd ul-Rahman Lomax]

At 09:08 PM 5/18/2013, George Sanders wrote:
>My problem with IRV: It's just another "majority wins" voting method
>(where a small majority can tyrannize a large minority). So, what's
>so special about going from 49.99999% to 50.00001%--did something
>'magic' just happen??

Yes. Something magic happened. That is, a threshold was crossed. We
need a threshold somewhere. Different thresholds are used for
different purposes, but for basic democratic decision-making process,
the *minimum* threshold is a majority. It's almost always desirable
to do better than that.

I don't find making a decision by a majority vote to be "tyranny."
Start thinking, George, in simple terms of how we, as a society, make
decisions. Think in terms of many decisions, not just one, and *stop*
thinking in terms of political parties with platforms, where issues
are heavily linked.

I've had experience with standard decision-making process, and it's
very simple, and there is no *tyranny* about it, unless somehow a
majority *has* decided to harass and oppress a minority. That does
happens, but it has nothing to do with the basic idea that *we* make
decisions by the preponderance of the votes, with majority being the
minimum level for any decision. Below that, we don't make a
collective decision.

Public elections by plurality are a glaring exception. They've been
thought necessary, but that was a misunderstanding of the possibilities.

Asset Voting can make certain decisions with functional *unanimity.*
That may seem totally crazy to those who have not experienced what
happens with Asset, and the possibilities for consensus in
decision-making process. Consensus is a goal that, in large groups,
can be difficult to attain, and my conclusion has been that it is a
group decision as to how far to go. I.e., how many cows have to come
home before we say it's enough?

The threshold for that decision is, I conclude, a majority. When
those who want to continue discussion are fewer than those who dont,
it is *oppressive and a form of tyranny* to demand that they
continue. It is, in fact, not an avoidance of the tyranny of the
majority, it is *minority rule*, and I have seen this play out in
groups unfortunate enough to have established consensus as a rule
*rather than as a highly desirable goal.*

Call it the tyranny of the past. (I.e., this situation arises when
the status quo favors a minority, and consensus is required for change.)

[Stephen Unger]

My responses are prefaced with ****

Steve
............

Stephen H. Unger
Professor Emeritus
Computer Science and Electrical Engineering
Columbia University
............

On Sat, 18 May 2013, Michael Ossipoff wrote:

>>Why is IRV being seriously considered here?

> Considered for what?

****Considered for use in elections

>There are 3 different positions on IRV that have been
> expressed here:

>
> 1. All-ouit opposition to IRV, without regard to conditioins.
>
> 2. FairVote representatives have spoken here too.
>
> 3. My claim that IRV and Approval each better than eachother, for
> different conditions.

****Obviously, I was referring to assertions such as #3, where Michael
has just confirmed that he considers IRV better than approval for
some conditions.

> It would be difficult or impossible to answer Steve's question
> without more detailed specification of what consideration he's
> referring to.

****My point is that the significant probability that an IRV election
will exhibit one of its weird behaviors should disqualify it for
consideration under all circumstances.

> On Saturday, May 18, 2013 12:12:01 PM UTC-4, Steve wrote:

> >I find it strange that there is so much discussion here in which
> >IRV (Instant Runoff Voting) is treated as tho it deserved
> >serious
> >consideration as a voting system.

> Steve is surely referring to the dedicatedly anti-IRV people who
were telling the reasons why they don't like IRV.

****The above has to be some sort of typo. Obviously I am referring
to those, such as Michael, who argue that IRV should be seriously
considered as a reasonable choice under certain
circumstances. (Which Michael is doing in this posting.)

> Funny, I didn't hear them considering IRV.What were they considering
> proposing it for?

****See Michael's item 3 above as one example.

> But, to use "consider" a little differently, surely you understand
> that your criticism of something >won't count for much if it's clear
> to people that you dont, give it fair and serious cnsideration. If
> you don't, then you're just engaging in knee-jerk assertion and
> monologue.

****I (and others in our group) have made numerous detailed arguments,
complete with examples, showing how IRV is capable of a number of
different kinds of pathological behavior, none of which are
exhibited by score or approval. My contention is that any system
capable of such behavior should not be considered as an option for
any election, regardless of any other properties that it may have.

> > There is much debate about the extent to which it favors
centrists or extremists

> That's an over-dramatic way of saying that it can fail to elect a
> CW, as can Approval.

****Here Michael has failed to appreciate that cases where approval
does not elect a CW (Condorcet winner) are examples of an
advantage, not a drawback. Approval will ALWAYS elect a candidate
that would have defeated all other candidates in 2-candidate
APPROVAL elections. Its results always make sense. (The same goes
for score voting.)

****Consider the following examples, where IRV and score/approval
disagree as to the winner. In the first case score/approval
chooses the CW and IRV does not. In the second case, IRV chooses
the CW and score/approval does not. In both cases, it is obvious
that the score/approval choices are correct.

****
Example 1.
IRV votes
# votes
4 B>A>C
3 C>A>B
2 A>B>C
A is the CW. B wins the IRV election in 2 rounds.
Score votes below are consistent with the above IRV
election. (Scores range from 0-9)
# votes A B C
4 8 9 0
3 8 0 9
2 9 1 0 Scores: A, B, C: 74, 38, 27. A wins
Approval election, also consistent with the IRV election.
# votes
4 AB
3 AC
2 A Approvals: A=9, B=4, C=1. A wins

****
Example 2.
IRV votes
# votes
4 B>A>C
2 C>A>B
3 A>B>C
A is the CW. A wins the IRV election in 2 rounds.
Scores votes consistent with the IRV election votes
# votes A B C
4 1 9 0
3 1 0 9
2 9 8 0 Scores: A, B, C: 39, 60, 18. B wins
Approval election consistent with the IRV and score elections
above.
# votes
4 B
3 C
2 AB Approvals: A=2, B=6, C=3. B wins
In these examples, the score and approval results, contrary to the IRV
results, make sense.

****It is important to understand the distinction between the CC
(Condorcet criterion--choosing a candidate (the Condorcet winner
or CW) that is ranked over ever other candidate) and the
consistency criterion: choosing a candidate that would beat every
the candidate in a 2-candidate election of the type
considered. E.g., in Example 2 above, A is the CW, but in the
score election, B outscores A 60-39. It is not a fault if CC is
violated, in favor of abiding by this consistency criterion, but
rather a favorable point. Score and approval never violate
consistency, while IRV often does.

> IRV can do so automatically, when no one intended to leave that CW
> unprotected, and so it can fairly be said that Approval is
> better-disposed
> to elect CWs than IRV is.

> When IRV elliminates a CW, it elects someone who receives the CW's
> transfers. It elects someone who is the next choice of the
> CW-preferrers.
> You conveniently forget that when you hysterically warn that iRV
> elects
> extremists.

****I never warned, "hysterically" or otherwise, "that IRV elects
extremists."

…

> Members of this group should be aware that every voting system fails
> criteria. In fact every voting system fails criteria that are
> obviously desirable.
 
> You give up one criterion in order to gain another one.  Members of
> this group should understand that.
 

****The various criteria appearing in our discussions are by no means
> of equal importance. Some are of minuscule value, and some, such as
> the CC are useless in that violating them in some cases, as shown
> above, is a positive behavior. Score and approval are remarkably
> free of violations of significant criteria, and those they do fail
> have to do with strategic issues that are often debatable.

>> Have people forgotten about the following IRV characteristics?

>> 1. If all votes are reversed, the same candidate might still
win. More generally, reversing every vote will very often not
lead to the order of the results being reversed.

 
> No one has answered my question about why that's important. How does
> it cause  strategy problem that could result in societal harm?

****Strategy is not the point here. Is it really necessary to present
reasons why it is an important failing for an election system to
sometimes rule that an election (where there are no ties) has
shown that the candidate MOST favored by the voters is also the
one LEAST favored by the voters? Can you imagine the general
outrage if an election for the governor of a state displayed such
behavior?

. . .

>> 2. Possible failure to elect a candidate that IRV elections
would show as beating every other candidate in a 2-candidate
election.

 
> Yes, IRV fails the Condorcet Criterion (CC).  Approval fails CC
> too. So does Score.

****See above for the difference.

**** I'm going to stop commenting at this point. It is obvious that
Michael is concerned only with "strategy" issues, which I regard
as of secondary importance as compared with matters of rational
interpretation of votes cast. If a voting system is so logically
noisy as to produce ridiculous results with non-negligible
probability, then I see no point in arguing about points of
strategy. Even the obviously important spoiler issue fades in
significance if the system is going to choose the winner with a
large degree of randomness.

[Dave Ketchum]

We get buried too deep in details to think clearly about the major
problems. So, trying:

Start the path where we all started, with Plurality.

First step should be Approval - an easy step that voters can make
easily, or still act, as voters, as if in Plurality if they do not
need, personally, to vote for more than one.
Big deal is that Primaries are not needed since Approval does
not need to avoid multiple clones, etc. This simplifies complications
of elections.

Next steps are toward more complex elections, such as:
Score and other uses of ratings. Doing ratings makes assigning
values to making each choice.
Condorcet and other uses of rankings. Doing ranking means
ranking each pair of candidates as to which is better (or equal), but
not able to or required to value the differences.
NOT IRV! IRV's use of ranking attracts with apparent
simplicity, but IRV often gets rejected once understood better. It
discards the candidates that are weakest according to each voter's top
ranking. This is a great idea for most elections - the exceptions
make the method discardable. I will try describing a subset that may
help complete this thread:
. We have what is, or at least looks like, a CW for candidate
"A". Either or both of the following will be seen:
. It's top level gets discarded as weakest.
. Discarding another makes it visible only at a lower level.
Whatever becomes visible this way is discardable since it cannot be
larger than what was discardable when seen at top level.
. Seeing that IRV too-often discards the CW sours all involved.

Dave Ketchum

[Michael Ossipoff]

I accidentally posted this message to the wrong thread. I'm now posting it to the right thread.

 

On Sat, 18 May 2013, Michael Ossipoff wrote:

>>Why is IRV being seriously considered here?

> Considered for what?

[quote]

****Considered for use in elections

[/quote]


Nonsense. There are elections of many kinds. There are many different
applications for voting systems, different voting situations. And
there are different conditions.

For example, I've defined "current conditions" and "Green scenario
conditions". For organizational voting, I've spoken of amicable and
inimical electorates.

And, of the frequent posters who've discussed IRV here, I'm the only
one who advocates IRV for any purpose. So, if Steve is referring to
anyone other than me, then he's mistaken if he thinks that they've
considered IRV for use in elecions of any kind.

> It would be difficult or impossible to answer Steve's question
> without more detailed specification of what consideration he's
> referring to.

[quote]

****My point is that the significant probability that an IRV election
will exhibit one of its weird behaviors should disqualify it for
consideration under all circumstances.

[/quote]

,,,,should, in Steve's mind.

I'm not saying that people's "shoulds" are wrong. But you're
completely wrong if you claim that your "should" is the right one,
without telling why you think so.

I've long been pointing out that, when people cite criteria, they
reallly need to say why they consider their criteria important or
necessary. In fact, I likely made that statement in the very posting
that Steve was replying to.

> On Saturday, May 18, 2013 12:12:01 PM UTC-4, Steve wrote:

> >I find it strange that there is so much discussion here in which
> >IRV (Instant Runoff Voting) is treated as tho it deserved
> >serious
> >consideration as a voting system.

> Steve is surely referring to the dedicatedly anti-IRV people who
were telling the reasons why they don't like IRV.

[quote]

****The above has to be some sort of typo. Obviously I am referring
to those, such as Michael, who argue that IRV should be seriously
considered as a reasonable choice under certain
circumstances. (Which Michael is doing in this posting.)

[/quote]

Steve is speaking of a plural number of people making such a claim.
Actually there's only one at this forum. But yes, I do make that claim.

But no, I don't merely argue that Approval should be seriously
considered as a reasonable choice under certain circumstances. I
assert that IRV is the best choice, or one of the best choices under
certain circumstances. (As in the Green scenario, if people aren't too
concerned about choosing the CW compromise, or letting non-MM voters
participate in choosing which MM-preferred candidate wins.),

Anyway, IRV isn't just being seriously considered as a reasonable
choice. IRV has been chosen, for the offered voting system, in the
platforms of at least four national political parties*. I agree with
and support that choice.

*Green Party U.S. (GPUS); Orignal, smaller, socialist Green Party
(G/GPUSA; Socialist Party USA (SPUSA); Reform Party

> But, to use "consider" a little differently, surely you understand
> that your criticism of something >won't count for much if it's clear
> to people that you dont, give it fair and serious cnsideration. If
> you don't, then you're just engaging in knee-jerk assertion and
> monologue.

[quote]

****I (and others in our group) have made numerous detailed arguments,
complete with examples, showing how IRV is capable of a number of
different kinds of pathological behavior, none of which are
exhibited by score or approval.

[/quote]

Different methods have different disadvantages. With the best rank
methods, it's possible to gain remarkable and powerful freedom from
strategy-need, for members of a muutual majority (and, with some of
those methods, good freedom from strategy need for all voters). Yes,
that means failing some monotonicity criteria, and similarly
strategically-irrelevant criteria.

[quote]

My contention is that any system
capable of such behavior should not be considered as an option for
any election, regardless of any other properties that it may have.

[/quote]

Steve's contention about his "should" isn't enough, of course. It's
necessary to tell _why_ he thinks that his criteria are important or
necessary.

> > There is much debate about the extent to which it favors
centrists or extremists

> That's an over-dramatic way of saying that it can fail to elect a
> CW, as can Approval.

[quote]

****Here Michael has failed to appreciate that cases where approval
does not elect a CW (Condorcet winner) are examples of an
advantage, not a drawback.

[/quote]

Not in elections where people care about strategy. Listen to Mr. Gilson.

Yes, if the election is u/a, then Approval strategy is simple: Approve
all of the acceptables, and none of the unacceptables. But that
doesn't let you choose among the acceptables. Any such choice must be
done, over subsequent elections, over the years, maybe taking an
addtional decade or more, before you find out what's the best you can
get, and finally achieve it.
.
Yes, Approval can be optimal in elections in which the alternatives
aren't different enough to make some of them odious to some voters.
Then there's less perception of strategy-need. Then people might be
satisfied to just approve what they like (I recommend that in
Approval). But, if conditions cause a perception of strategy-need, or
if you have strong preferences among your acceptable set, and don't
want to wait years and years to finally elect the best you can, then
you want IRV, AIRV, Woodall, Benham, Schwartz Woodall, MM-Woodall, or
MM-Benham instead of Approval or Score.

However, I assure Steve that, in the Green scenario, there would
surely soon be a national referndum or initiative to choose a national
voting system (There will be national initiatives and referenda, and
the Constitution could be amended to establish a national voting
system). In that initiative or referendum Steve would be welcome to
nominate Approval, and rank it in 1st place on the IRV ballot that
would be used in that Green America referendum or initiative. But,
Steve, don't forget that you'd have no reason to not rank as many
methods as you want to, because IRV meets LNHa.

You know, just in case Approval doesn't win.

[quote]

Approval will ALWAYS elect a candidate
that would have defeated all other candidates in 2-candidate
APPROVAL elections. Its results always make sense. (The same goes
for score voting.)

[/quote]

No, not really. The candidatre who'd beat each of the others in
2-candidate Approval elections would be the CW. ...if voters in those
2-way Approval elections vote in their obvious best interest. Likewise
for Score.

[quote]

****Consider the following examples, where IRV and score/approval
disagree as to the winner. In the first case score/approval
chooses the CW and IRV does not. In the second case, IRV chooses
the CW and score/approval does not. In both cases, it is obvious
that the score/approval choices are correct.

****
Example 1.
IRV votes
# votes
4 B>A>C
3 C>A>B
2 A>B>C
A is the CW. B wins the IRV election in 2 rounds.
Score votes below are consistent with the above IRV
election. (Scores range from 0-9)

[/quote]

6 voters rank A and B over C. That's a voted mutual majority. The
winner comes from that majority-preferred set. Furthermore, to
achieve that, the A voters and the B voters didn't have to do other
than rank sinicerely among {A,B}.

In your example, IRV has shown its sincerely-achieved majority-rule
power for a mutual majority.

Thanks for the example.

A is the CW, and A didn't win. The C voters weren't asked which
MM-preferred candidate to elect. They
weren't members of the mutual majority. Why should they decide which
MM-peferred candidate should win?

There's no ethical reason why it's necessary to include the non-MM C
voters in that choice.

Yes, there's a practical reason: The A voters and the C voters could
combine to make a dis-satisfied majority who would insist on replacing
IRV with Benham or (preferrably) Woodall. But that's fine. They can do
that via an initiative or referendum, in a Green U.S. government.



[quote]

[/quote]

What Steve believes "make sense" is a subjective matter of Steve's
opinion. Yes, Approval and Score better protect a CW, when people want
to protect hir.

That comes at a price. In Score and in Approval, the voters had a
strategic decision to make. In IRV, the A voters and the B voters
didn't have to concern themselves with strategy. And Woodall or Benham
would have automatically elected the CW, candidate A, while retaining
IRV's powerful combination of MMC and no chicken dilemma.

[quote]

****It is important to understand the distinction between the CC
(Condorcet criterion--choosing a candidate (the Condorcet winner
or CW) that is ranked over ever other candidate) and the
consistency criterion: choosing a candidate that would beat every
the candidate in a 2-candidate election of the type
considered.

[/quote]

The Consistency Criterion says that if X wins when only the district 1
votes are counted, and X wins when only the district 2 votes are
counted, then Xshould win when both the district 1 and district 2
votes are counted.

[end of Consistency Criterion definition]

That criterion is met by Approval and Score, but not by any Condorcet
method, and not be IRV. It, and the monotonicity criteria, and
Reversal Symmetry, could be
classified into a broad category of criteria called "consistency
criteria". When one of them is violated, Steve will say that something
"weird" has been done. ...but not something that causes a strategy
problem that could cause societal harm.

> Members of this group should be aware that every voting system fails
> criteria. In fact every voting system fails criteria that are
> obviously desirable.

> You give up one criterion in order to gain another one. Members of
> this group should understand that.

[quote]

****The various criteria appearing in our discussions are by no means
> of equal importance. Some are of minuscule value, and some, such as
> the CC are useless in that violating them in some cases, as shown
> above, is a positive behavior.

[/quote]

I agree that sometimes it's ok if CC is violated. That's why I like
IRV. That's why I like MM-Woodall and MM-Benham.

But,when CC is violated, there will be a majority who would have
preferred electing the CW. A dis-satisfied majority could insist on
enacting a CC-complying method. That's why I've suggested that Woodall
or Benham might eventually replace IRV. ....or Approval.

[quote]

>Score and approval are remarkably
> free of violations of significant criteria

[/quote]

...by steve's standards regarding which criteria are significant.

Approval and Score fail MMC and they both have the chicken dilemma.

I've told why the chicken dilemma wouldn't be a _problem_ in Approval
and Score (at least not if you don't mind a little extra work, taking
maybe an extra decade or two, before you get the best result you can). I call
the chicken dilemma a nuisance. But IRV, Woodall and Benham don't have
that nuisance. Unlike Approval or Score, IRV, Woodall and Benham will
immediately enforce majority rule for a MM, without that MM using any
strategy.

> and those they do fail
> have to do with strategic issues that are often debatable.

No doubt everything is debatable, as is evident from these forums. But
Steve seems to want to implly that strategy criteria are perhaps more
"debatable", and less "significant" than his various consistency
criteria (Mono-Raise, Participation, Mono-Add-Top, Consistency,
Reversal Symmetry). That's Steve's opinion. Ask him how those
criteria measure for something that would result in societal harm, to
the degree that strategy problems and failure of majority rule can
result in societal damage.

>> Have people forgotten about the following IRV characteristics?

>> 1. If all votes are reversed, the same candidate might still
win. More generally, reversing every vote will very often not
lead to the order of the results being reversed.


> No one has answered my question about why that's important. How does
> it cause strategy problem that could result in societal harm?

[quote]

****Strategy is not the point here.

[/quote]

No, not to Steve.

[quote]

Is it really necessary to present
reasons why it is an important failing for an election system to
sometimes rule that an election (where there are no ties) has
shown that the candidate MOST favored by the voters is also the

one LEAST favored by the voters? [/quote]

No, to a True Believer no reasons need be presented.

[quote]

Can you imagine the general
outrage if an election for the governor of a state displayed such
behavior?

[/quote]

Well perhaps people can imagine Steve trying to stir up such outrage.

In IRV, if a set of candidates is ranked over the others by a
majority, and you reverse the rankings, then that set will be ranked
_below_ the others by a majority.
Before the reversal, one of them would win. But not after the reversal.

I don't know if IRV violates Reversal Symmetry, but, if it does, it
will turn out to be an entirely unimportant case.

Notice that Steve doesn't have an example of IRV failing Reversal
Symmetry, and so we don't know whether or not he thinks that the
example is important, or can tell why it's important.

>> 2. Possible failure to elect a candidate that IRV elections
would show as beating every other candidate in a 2-candidate
election.


> Yes, IRV fails the Condorcet Criterion (CC). Approval fails CC
> too. So does Score.


****See above for the difference.

I've always acknowledged that Approval better protects a CW that
people want to protect. And in Approval they protect hir by strategy.
Approval and Score, compared to IRV, are to be noted for their need
for strategic judgments, which someone (maybe Steve) referred to as
"agonizing". In IRV, a MM needn't agonize in order to ensure victory
for their MM. They need only rank sincerely.

If you want to reliably elect CWs, in the Green scenario, then use
Woodall or Benhlam, not Approval or Score.

Under current conditions, ICT and (better yet) Symmetrical ICT meet (a
reasonable and justifiable version of) the Condorcet Criterion.
Approval and Score don't.

[quote]

**** I'm going to stop commenting at this point. It is obvious that
Michael is concerned only with "strategy" issues

[/quote]

Yes. I've amply told why.

[quote]

, which I regard
as of secondary importance as compared with matters of rational
interpretation of votes cast.

[/quote]

Steve is referring to consistency criteria, such as the monotonicity
criteria (Mono-Raise, Participation, Mono-Add-Top,
Mono-Add-Unique-Top, Consistency, and Reversal Symmetry.

Strategy problems are associated with consistent and reliably
predictable societal harm. Consistency
criteria failures are not.

[quote]

. Even the obviously important spoiler issue fades in
significance if the system is going to choose the winner with a
large degree of randomness.

[/quote]

The "obviously important spoiler issue", under current conditions,
guarantees perpetual rule by the same two unliked parties (actually
two right wings of one party).

Yes, any good rank method will do some capricious things (referred to
by Steve as "random"). That's the price for the powerful strategic
guarantees offered by the better rank methods, and the resulting
societal benefits, such as the much more immediate and rapid societal
improvement resulting from automatic majority rule enforcement, at
least for a MM (but for everyone, if the method meets CC, as do
Woodall and Benham).

Michael Ossipoff

- show quoted text -

Click here to Reply or Forward

Show trimmed content

me (Michael Ossipoff change)

11:03 PM (4 minutes ago)

Other recipients:

On Thursday, May 23, 2013 6:45:58 PM UTC-4, Dale Sheldon-Hess wrote: