Revisiting RangeVoting's page on STAR Voting

[Mark Frohnmayer]

Hey all, just wanted to encourage folks to take another look at Warren's
page about STAR Voting (https://rangevoting.org/StarVoting.html).

The reason I bring this up is that Schulze used Warren's page as a source on Wikipedia to reinforce the notion that STAR does not satisfy the monotonicity criterion. I believe that conclusion to be incorrect: by the definitions of the monotonicity criterion on CES, Warren's own site, and Wikipedia itself, STAR undeniably satisfies that criterion:  increasing support for a candidate in STAR cannot cause that candidate to lose, nor can decreasing support for a candidate cause that candidate to win. Warren's examples claiming STAR is non-monotonic rely on changing the level of support for other candidates, and therefore do not satisfy the actual accepted definition of the monotonicity criterion (copied from Warren's own site):

Monotonicity is the property of a voting system that both:

- If somebody increases their vote for candidate C (leaving the rest of their vote unchanged) that should not worsen C's chances of winning the election.

- If somebody decreases their vote for candidate B (leaving the rest of their vote unchanged) that should not improve B's chances of winning the election.

The errors in the analysis don't stop there. In Warren's section labeled "Voter honesty," he concludes "Voters are motivated to lie in their STAR votes," yet he provides no actionable heuristic that voters can employ to decide when such a dishonest vote is warranted.

STAR Voting will be on the ballot in Lane County, Oregon this November. The only "opponents" who have surfaced so far are supporters of other voting systems such as FairVote (see http://equal.vote/fv) and Smith who promulgate FUD that is then picked up by others to diminish support for the effort here. We are glad that Warren at least states his bias in the first paragraph ("I'm biased against it and dubious"), but we are disappointed that others have chosen to use his admittedly biased draft objections as a source for further misdirection.

Any feedback welcome.

Cheers,
Mark

Mark Frohnmayer
Founder, Equal Vote Coalition
Co-Chief Petitioner, STAR Voting for Lane County

[NoIRV]

Pop quiz!
When it comes to voting reform, ________ is not a reliable source.
A. (Un)FairVote
B. Wikipedia
C. Both A and B
D. None of the above

(The answer is C.)

[NoIRV]

> Pop quiz!
> When it comes to voting reform, ________ is not a reliable source.
> A. (Un)FairVote
> B. Wikipedia
> C. Both A and B
> D. None of the above
>
> (The answer is C.)

Although, in theory one of us could edit out the lies on Wikipedia. (I think I remember IRV being stated to "eliminate the spoiler effect" andor "be immune to strategy" on Wikipedia.)

[Warren D Smith]

I thought my STAR page explained what it meant, and also explained why
STAR sort of points out limitations of the old monotonicity definition
wording, and really should not be regarded as monotonic under
a new better wording.

This was not an "error" by me. It was an "attempt to delude people
into thinking
there was an error" by you.

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

[Clay S]

Warren,

Can you describe any general strategy that a voter can follow to get a better result than honest ballot?

Ut is pretty astonishing to me how unsupportive you are of the only ballot initiative in the history of the United States to put a Score Voting ballot in front of voters.

[Andy Jennings]

Warren,

There's a typo on that page.  Where it says:

Then the score-sums are C77, B72, A47 

Should be B56.  Doesn't change any results, though.

That whole last section does seem a bit of a reach, to me.

In the first example, when those 16 voters lower A one point, that changes the winner from C to B.  Congratulations, you've proved that STAR is not IIA.  It doesn't matter whether they raise C one point or not.  I'm not sure why you feel the need to say they also raise C by some epsilon so you can call it "non-monotonicity".

In the second example, when those two voters lower A four points, it changes the winner from B to C.  Congratulations, you've proved that STAR is not IIA.  It doesn't matter whether they raise B four points or not.  Again, I'm not sure why you feel the need to say they also raise B by some epsilon so you can call it "non-monotonicity".

~ Andy

--
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 electionscience+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Jameson Quinn]

You're right; it wasn't an error. It was an attempt to redefine monotonicity to something you think is more important. You may or may not be right, but I still think that "monotonicity" should not be the word for your new concept.

STAR is monotonic. 

[Jameson Quinn]

I agree with Andy's point... and also note that this discussion seems to straddle multiple lists, which is confusing to me.

[Sara Wolf]

Hi Warren, 

What do you think about the Equal Vote 5 pillars to evaluate voting methods? Accuracy, Equality, Honesty, Expressiveness, Simplicity? 

I believe that the pass/fail criteria have benefit, but especially in cases where definitions are in flux or in this case where you don't like the existing definition it gets problematic. The end goal here is to evaluate how good a system is overall, and passing or failing x criterion is only one consideration. At some point we do need to zoom out and get excited that we have an expressive cardinal system on the ballot with a ton of people excited about it. Moreover that system is topping the charts for VSE, does not incentivize strategic voting, and would likely have a great Bayesian Regret score as well. It provides an equally weighted vote. Nothing is perfect, but that is a pretty strong and well rounded package. 

I know you are a mathemetician and scientist, not a politician or an activist, but this is our opportunity to dramatically improve Oregon elections and to give voters a real alternative to IRV. We are asking for your help. Please consider editing your article or at least requesting that Wikipedia not site it as you admit that you are biased against STAR. (Thank you for that transparency!) I was frankly shocked that you wrote that article with that statement at the beginning. I'd been specifically citing your work to people as an unbiased source. That's the kind of review we need right now. Please! 

Anything you can do to help us here would be extremely appreciated. 

[Warren D Smith]

On 8/16/18, Sara Wolf <sa...@equal.vote> wrote:
> Hi Warren,
>
> What do you think about the Equal Vote 5 pillars to evaluate voting
> methods? Accuracy, Equality, Honesty, Expressiveness, Simplicity?

--can you refer me to a writeup you have in mind?

I've seen some stuff which struck me as annoyingly oversimplified
and slanted, which this may or may not be.

STAR seems superior to plurality, but I'm dubious it is better than (in fact
it seems worse than) plain range or range+sep.runoff in terms of
Accuracy, Simplicity, Honesty;
it is the same as range and worse than range+runoff in terms of Expressiveness;
and I don't know what you meant here by Equality but in all three systems
I guess all voters are entirely "equal".

The whole STAR thing annoyed me from inception to now. It was about
solving a non-problem while increasing complexity and spewing baloney;
and there seems no evidence
it is more enactible, in fact its purveyor Mark Frohnmayer massively
overestimated its enactibility to me from day one, thus ruining his credibility
with me on that question. Its greater complexity strikes me as a
worry re enactibility, since it looks to me based on limited evidence that
every extra bit complex you go, the less enactible voting systems get.
Even the name "STAR" annoys me, e.g. STIR seeming
more accurate descriptor. Its other inventor Rob Richie seems to have largely
disowned it. Its third inventor Clay Shentrup, I have no idea.

But, I think STAR is superior to plain plurality.

[Sara Wolf]

"What do you think about the Equal Vote 5 pillars to evaluate voting methods? Accuracy, Equality, Honesty, Expressiveness, Simplicity? - can you refer me to a writeup you have in mind?"

This presentation I gave to Sightline is probably the best. I tried to present the info in a way that had some real substance, but would still be digestible for a lay person, as many of the people making recommendations on voting reform have not studied this stuff. I hope you can forgive me for presenting information at times to lay people that is admittedly oversimplified and intended to be persuasive. It's a campaign after all. But I hope that the information itself is still essentially accurate and of value. 
https://www.starvoting.us/sightline

The 5 pillars are not a substitute for criterion that are mathematical proofs, but rather are a larger framework that we can view systems through. The graphics/grades used are not hard science and are absolutely subjective, (as the disclaimer says,) especially since a few of these pillars are multi-faceted. Regardless, they are a good tool for looking at the issue as a whole.

For example when we evaluate accuracy, even if we disagree on VSE vs. Bayesian Regret, vs. Condorcet, etc. we can agree that we want elections to be accurate. FairVote has managed to reject every single measure for accuracy, while completely disregarding it as a consideration at all. These 5 pillars are an attempt to reframe the conversation so critical considerations don't fall through the cracks. Depending on our preferred measures and considerations we may not be able to agree on an exact grade that each system should receive for "Accuracy", but we can say that by any measure Ranked Pairs is more accurate than IRV, for example, according to the information we have. Short of absolute proof we can still make evidence based recommendations in an ever evolving field. https://www.starvoting.us/accuracy

"STAR seems superior to plurality, but I'm dubious it is better than (in fact it seems worse than) plain range or range+sep.runoff in terms of Accuracy, Simplicity, Honesty;" 

The answer to those questions lies in real world data on strategic voting patterns for all three systems, which we don't have yet. All the more reason to support each others work and see all three used in real world elections. I promise that I will be supportive of proposals for any of those three options over plurality or IRV! I am positive that all three would be a HUGE improvement over our current system. I agree with you that the strategic voting concern has been overblown and I'd love to see data come through to disprove it or show that it's impact is limited.

Still, the reality here is that it is a major concern for many and was a deal breaker here in Oregon. Top-2 is also a deal breaker for 3rd parties. Here in Oregon the cardinal/score voting community is united in advocating for STAR Voting (I can't think of any exceptions.) Our core team is made up of people who 2 years ago were advocates for Score, Approval, RCV, and Condorcet. This really is a proposal that directly addresses the criticisms against each of those systems. It's a compromise, but it also just has a ton to offer on it's own merits. 

"I don't know what you meant here by Equality but in all three systems I guess all voters are entirely "equal"."

Yes. They all pass the "Test of Balance" (aka Equality Criterion. Is that a recognized term?) https://www.starvoting.us/equal_vote

"there seems no evidence it is more enactable,"

We have broad support across Oregon for this, we've ran 2 ballot initiatives so far, and now have this on the ballot for voters in Lane County to adopt on Nov 6th. Score Voting fans got on board here because they like it and also because it does have much broader support and enactability. The runoff does change the strategic incentives and that's a huge deal to many, even people who have never heard of FairVote. 

STAR Voting has no open opposition, but we do have FairVote putting out FUD while claiming neutrality. Your article is cited against us on wikipedia, which is a top site that comes up when you google STAR Voting. You may not have intended it for that audience, but your piece is now a leading source cited to encourage people to stick with Plurality. I know you don't want that!

 "Even the name "STAR" annoys me." 

Marketing is indeed annoying! (At times so is Mark.) I hate that we have to jump through those kinds of hoops to have a chance at enacting good policy, but that's politics. These decisions should be based on the merits of the idea itself, but we have a long way to go. You wouldn't believe some of the reasons politicians and lobbyists cite for their policy stances! It's deeply depressing, to say the least. 

"Its other inventor Rob Richie seems to have largely disowned it. Its third inventor Clay Shentrup, I have no idea."

Rob is actively opposed to all voting reforms besides RCV. Clay's new group Counted is part of the Equal Vote Coalition (we're working to make that more official) and we support each others work. He likes STAR and Score and hopes to see both passed for real world elections. Clay has offered his official endorsement for our initiative. We would love yours as well. 

"But, I think STAR is superior to plain plurality."

Thank you for that. Do you think it's superior to IRV too? 

[Warren D Smith]

> Still, the reality here is that it is a major concern for many and was a
> deal breaker here in Oregon. Top-2 is also a deal breaker for 3rd parties.

--strange considering that top-2 runoff causes 3-party domination, or
more than 2 parties anyhow, in most countries that use it; but IRV
causes 2-party domination in every country that uses it. (And despite
that the USA Green Party endorses IRV!)
This proves the USA Green's problem is ignorance.

It is also interesting that, e.g. honeybees use plain range (not STAR...)
which is both testimony for it, and also (you might have thought) an attraction
for Greens.
https://rangevoting.org/ApisMellifera.html

> We have broad support across Oregon for this,

--the proof of that is in the pudding. If any. I suspect you would
have had broader support for plain range. There is limited evidence
about what systems voters want, see
https://rangevoting.org/WhatVotersWant.html
https://rangevoting.org/CanadaVoteHistory.html
but there seems to be a trend in that evidence that they do not want,
and/or can easily be persuaded to be fatally suspicious of,
complexity. It takes very little complexity before there is enough
grist for the FUD-mill to doom its enaction. I gave plenty of nasty
examples of STAR pathologies in my piece which make it clear it is a
lot more complex and goofy than one might naively imagine.

Also, if succeed in Oregon you then serve as role model for
other states & countries to emulate. It then would be better to be a
better role model.

> STAR Voting has no open opposition, but we do have FairVote putting out FUD
> while claiming neutrality. Your article is cited against us on wikipedia,
> which is a top site that comes up when you google STAR Voting. You may not
> have intended it for that audience, but your piece is now a leading source
> cited to encourage people to stick with Plurality. I know you don't want
> that!

--well, actually "my piece" clearly says STAR is better than plain plurality.
So, if it is "now a leading source

cited to encourage people to stick with Plurality"

again the problem is ignorance.

> Thank you for that. Do you think it's superior to IRV too?

--Yes. STAR is less voter-error prone than IRV, more expressive,
doesn't ignore a large fraction of what voters say, is
comparably complicated to count, and better in terms
of NESD-type criteria so that it is not
as bad as IRV in terms of risk of causing 2-party domination.

[Clay S]

On Friday, August 17, 2018 at 7:19:01 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

It is also interesting that, e.g. honeybees use plain range (not STAR...)

Bees are eusocial, so strategic behavior is clearly less of a concern.

It is interesting that after all these years, you have yet to show a generally advisable STAR strategy that's better than honesty. One could call that a pretty strong endorsement of STAR Voting.

> It takes very little complexity before there is enough grist for the FUD-mill to doom its enaction.

But it's substantially simpler than IRV, which has already been adopted by an entire state.

[Sara Wolf]

top-2 runoff causes 3-party domination, or

more than 2 parties anyhow, in most countries that use it; but IRV

causes 2-party domination in every country that uses it.  (And despite

that the USA Green Party endorses IRV!)

This proves the USA Green's problem is ignorance.

 

Ignorance is definitely the problem and it's other organizations too. There are understandable reasons to not get on board, but they are not the ones cited. It's all propaganda. I really expected the Green Party to be the first on board here, and a few of their/our members have been, but the leadership's position can only be described as divisive and defensive. That seems to be changing though. The honeybee analogy is great! I'll bring that up. 

I'm sure voters would prefer plain score at a glance, but the reality is that unless voting advocates themselves are convinced and willing to do the years of hard work required to get reform on the ballot it will never come to that. Here in Oregon we had a 3 month debate and discussion before we voted on which voting method to work on. I and many started in the IRV camp and got won over. At the beginning of that process I'd say most preferred IRV, a handful preferred Approval or Score, and Mark and his friends presented SRV. 

At the end of that long process of debating the options, roughly 60% preferred STAR and 40% preferred IRV. The Score and Approval people still like those systems, and most people like Score more than the rest, but like that STAR is more accurate and strategy resilient. When I go out and do presentations to neighborhood groups people often cut me off midway and explain that they would bullet vote. (Even if they've never heard of that term many people think of the objection on their own. :/ ) I then explain the runoff and they are satisfied. The runoff also has a huge selling point of ensuring that your vote isn't wasted. Even if your favorite can't win, if you had a preference, your full vote will go to the finalist you preferred. People love that. They are SO sick of wasting their vote or being bullied into voting lesser evil. 

--well, actually "my piece" clearly says STAR is better than plain plurality. So, if it is "now a leading source cited to encourage people to stick with Plurality" again the problem is ignorance.

Yeah. I get what you were saying, but I don't think most can follow all of that science and just go with the initial bit which makes it clear you don't like it. Any editing you could do to make it clear that this is preferred over both IRV and Plurality, would be really appreciated, if you'd consider doing that. 

[Warren D Smith]

On 8/18/18, Clay S <cshe...@gmail.com> wrote:
> On Friday, August 17, 2018 at 7:19:01 PM UTC-7, Warren D. Smith (CRV
> cofounder, http://RangeVoting.org) wrote:
>>
>> It is also interesting that, e.g. honeybees use plain range (not STAR...)
>>
>>
>
> Bees are eusocial, so strategic behavior is clearly less of a concern.

--at an earlier time, when these behaviors and indeed the entire concept
of eusociality were evolving, bees must have been less cooperative and
more strategy-incentivized than now.

Let me make an analogy you probably have more intuition about...
consider the cells in your body. They are cooperating, you hope.
(Although in the earliest days, when multicellularity was a "new idea,"
the cooperation must have been much poorer.)
Only a tiny subset of your cells, get to reproduce to create a new human.
The rest are there to support them.

Now even today, you are at considerable risk for cancer. It is the 2nd-leading
cause of death in the USA, and almost in first place, and that has been true for
every one of the last 75 years. Cancer is cells deciding
not to cooperate, and to reproduce a lot in violation of the Cooperative Plan.
This leads temporarily to great "success" for those cells.

Also, other kinds of violations/failures of the Cooperative Plan are
responsible for auto-immune diseases. And all the bacteria in your gut
can be regarded as part of your biology
too ("microbiome") and they too have to obey rules and sometimes misbehave
mostly due to acting in "their own interests" rather than as part of
the Cooperative Plan.
Also, inside your DNA are "greedy" code-segments such as "ALU" and retroviruses
that are "out for themselves" and are in fact deleterious to you as a whole.
If we throw those in too, in addition to cancer, then probably
Failures of Cooperation
in sum are the leading cause of USA death.

So obviously, the Cooperative System of cells that is you, is still Darwinianly
evolving, and the top force acting to push that, remains, even today,
the defects
in the present Cooperative Plan, most of which, even today, are related to cells
and/or DNA subsegments acting "in their own interests" (or what they
naively "think" are their own interests).

And all that is despite the fact, that multicellularity is way, way older and
more evolved than honeybees.

> It is interesting that after all these years, you have yet to show a
> generally advisable STAR strategy that's better than honesty. One could

> call *that* a pretty strong endorsement of STAR Voting.

--well, this non-action on my part probably has a large relationship
to the fact that I never tried, during "all those years." Nor was I
ever asked.
And has anybody else ever tried? I'd suspect if I worked on it for few
days I could do that inside some simulated world.
And not just for STAR, actually for most voting systems.

Conitzer & Sandholm proposed a general purpose voting strategy which
seems to be better than honesty for most voting systems out there, although
not necessarily all of them. It is: identify the two
most-likely-seeming-to-win of
the N options. Artificially score them top & bottom (honest order
among those 2).
Now do something unspecified (may depend on the system) about the remaining
N-2 options. If everybody but you vote honestly, and you do this, then in most
voting systems out there, that is better for you on average than being honest.

Furthermore, it looks like a large percentage of real voters actually
use this strategy.
Unfortunately that causes bad effects, for example in plurality and
IRV voting this causes
2-party domination. But with approval and especially score voting 2PD
is not a forced consequence, not even with 100% of voters using this
strategy, even in its strongest
form where the top and bottom are demanded to be sole-top and sole-bottom.
For STAR the effects are intermediate: if 3-candidate elections then
it (at least in
the strong form) forces 2-party domination, but
STAR can avoid forcing 2PD in 4-option elections.

But in systems (including STAR?) involving a 2-man runoff, the
perhaps-more-natural
and/or better analogue of this strategy might be to identify the two
candidates seeming-most-likely to come in 2nd and 3rd, then
artificially vote them TOP & BOTTOM, then do something unspecified
(may depend on the system) about the remaining N-2 options.
That strategy if widely adopted might mess things up even worse. In
fact there is some suspicion it has messed up France.

>> It takes very little complexity before there is enough grist for the
> FUD-mill to doom its enaction.
>
> But it's substantially simpler than IRV, which has already been adopted by
> an entire state.

--whether STAR is simpler than IRV, is arguable.

If you are referring to Maine, that was (a) a huge fuckup and
a massive embarrassment for voting-reform activists everyplace,
(b) only 2 alternatives were offered to the poor voters, IRV & plurality,
both bad, (c) I think the majority of historical cases where those 2
alternatives
were the only ones offerred, IRV lost, or was later repealed.

[Jameson Quinn]

As far as I can tell, Warren, all of your points in this email amount to poking holes in other people's arguments while suggesting that you could, if you wanted to, come up with arguments against STAR. For instance, you point out that incentives are not always perfectly aligned for bees, without actually suggesting any mechanism by which this could lead them to vote strategically ("this potential nest is one where I or my full siblings could prosper and reproduce better than our nestmates, as compared to that potential nest" seems to me like knowledge no bee could have). 

Do you think that's the best thing you could be doing in the current context? 

STAR is on the ballot. I'm sure that many or most of us, including me, could each come up with something else that they think would be even better. But those other ideas are not options for Lane County; they will be deciding between plurality and STAR. Arguments which are not relevant to that decision are at best distractions, and in many cases actively counterproductive, right now. There's still a place for bigger-picture thinking, but even when it's appropriate, it should be clearly labeled as such, and preceded by a clear statement or link relevant to the current situation.

Warren, if STAR voting passes, a non-negligible part of the credit will go to your historic work on (what's now called) score voting. And I don't believe that Lane County, OR is the end of the road here. In your work today, do you want to help that important accomplishment, or not?

[Clay S]

On Saturday, August 18, 2018 at 6:36:53 AM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

--at an earlier time, when these behaviors and indeed the entire concept
of eusociality were evolving, bees must have been less cooperative and
more strategy-incentivized than now.

Worker bees cannot be strategic—they are sterile. They can only help copies of their genes in their hive-mates. There just inherently can't be an incentive for strategy. It would be as if a voter's satisfaction with an election outcome was the average of all other voters'.

Caveat: yes, of course, their individual genes can be strategic, in helping hive mates that possess them more than hive mates that don't posses them. But I don't think there was a historical point where bees were voting but were more more strategy-incentivized than now.

Cancer is cells deciding
not to cooperate, and to reproduce a lot in violation of the Cooperative Plan.
This leads temporarily to great "success" for those cells.

They don't succeed. They don't help themselves reproduce. In fact by harming the host, they are less successful. I don't think genes which cause cancer are being "strategic".

And all the bacteria in your gut
can be regarded as part of your biology
too ("microbiome") and they too have to obey rules and sometimes misbehave
mostly due to acting in "their own interests" rather than as part of
the Cooperative Plan.

That's different. Individual bacteria can reproduce.

 

Also, inside your DNA are "greedy" code-segments such as "ALU" and retroviruses
that are "out for themselves"

All of your genes are out for themselves.

And has anybody else ever tried?  I'd suspect if I worked on it for few
days I could do that inside some simulated world.
And not just for STAR, actually for most voting systems.

That would really help your criticisms of STAR Voting. Strange that you've not done it.

--whether STAR is simpler than IRV, is arguable.

In what way? Ballot spoilage, precinct summability, two rounds vs. n. STAR is simpler in every way I can think of.

[Warren D Smith]

On 8/18/18, Jameson Quinn <jameso...@gmail.com> wrote:
> As far as I can tell, Warren, all of your points in this email amount to
> poking holes in other people's arguments while suggesting that you could,
> if you wanted to, come up with arguments against STAR. For instance, you
> point out that incentives are not always perfectly aligned for bees,
> without actually suggesting any mechanism by which this could lead them to
> vote strategically ("this potential nest is one where I or my full siblings
> could prosper and reproduce better than our nestmates, as compared to that
> potential nest" seems to me like knowledge no bee could have).

--you've lost me. Is there something you are trying to argue for?

> Do you think that's the best thing you could be doing in the current
> context?

--I was asked my opinion. I gave it.
Do you think whatever you are writing is the best thing you could
possibly be doing?

[Warren D Smith]

On 8/18/18, Clay S <cshe...@gmail.com> wrote:

> On Saturday, August 18, 2018 at 6:36:53 AM UTC-7, Warren D. Smith (CRV
> cofounder, http://RangeVoting.org) wrote:
>>
>> --at an earlier time, when these behaviors and indeed the entire concept
>> of eusociality were evolving, bees must have been less cooperative and
>> more strategy-incentivized than now.
>>
>
> Worker bees cannot be strategic—they are sterile.

--actually a small percentage are fertile. This already was
discussed on the CRV page years ago...

> They don't succeed. They don't help themselves reproduce. In fact by

> harming the host, they are *less* successful. I don't think genes which

> cause cancer are being "strategic".

--congratulations, you have just proved cancer does not exist.
Unfortunately it not only exists, it is the 2nd leading cause of death.
Even now after billions of years.
This contradiction suggests something is wrong with your thinking here.
Maybe you should think more. I mean, your life is at stake.

> And has anybody else ever tried? I'd suspect if I worked on it for few
>> days I could do that inside some simulated world.
>> And not just for STAR, actually for most voting systems.
>>
>
> That would really help your criticisms of STAR Voting. Strange that you've
> not done it.

--why should I? I do not work for you. It is not the goal of my
life to focus on
what logically is an obscure offshoot. Also if I did, then you'd just
whine more
rather than actually being impressed as you imply here.

I daresay there are people accusing me of being "strange" by not focusing
more on showing the earth is round.

But ok, the Gibbard-Satterthwaite theorem goes a long way toward proving that
honesty is not best strategy for STAR. Which you should have already
known and hence never even raised this asinine "debate."


> In what way? Ballot spoilage, precinct summability, two rounds vs. *n*.

> STAR is simpler in every way I can think of.

--IRV elections are pretty much a subset of STAR elections at least
in the 3-option case.
Subsets are simpler than supersets, I would think.

[Sara Wolf]

I'd like to present a few more criteria for a great voting system. These are areas where I personally find STAR to be extremely compelling, yet which aren't usually focused on for some reason. (Control the framing control the conversation.) Election science will continue to prop up bad systems when bad criteria are prioritized over more important ones. Bringing criteria that directly speak to the goals of everyday voters would certainly be a boon to election science. Obviously advocates for any system could cherry pick criteria they pass and elevate them, as FairVote has done, but I believe these criteria listed below are desirable across the board for the next level of voting innovation. I know there is a way to word each of these so that STAR passes and I'm not sure I got that quite right here, but regardless of the details, STAR does a great job at each of these goals. 

1. Equality Criterion: The weight and worth of the citizens' votes is equal if the voter chooses to use the full weight of their vote. i.e. Any way that I fill out my ballot you can fill yours out in an equal and opposite way.

2. The Wasted Vote Criterion: A voter cannot waste their vote by voting their conscience. ie. even if your favorite or favorites cannot win, your honest vote can always help you get better representation than if you had not voted at all. (An honest vote contains a voters honest preference order and degree of support.)

3. Lesser-Evil Criterion: It is never a good strategy to rank or rate your lesser evil above your favorite. 

4. Consensus Criterion: If the ballots submitted in a given election contain information that shows that there is a candidate which has support from the entire electorate, the voting system must consider that information when selecting the winner that best represents the electorate. (This is the inverse of LNH)

5. Honest Incentive Criterion: Dishonest strategic voting, where a voter insincerely alters their preference order between candidates is not incentivized. ie. that strategy is more likely to backfire than it is to help them overall. 

I also would love to see the adoption of a walked back standard for election criteria as a whole. Pass/Fail is nice for mathematical proofs, but for instance there should be some sort of recognition if a system does a great job at both FB and LNH, while not getting 100% on either. Would that be an Election Criteria Standard where a passing grade is 90% or 95%, but 100% is recognized as the gold standard? Can that kind of high but not perfect standard be objectively evaluated using simulated elections based on relative frequency of real world election types? 

Lastly, in light of what Jameson said above I'd like to propose a pact that in light of the real world implications of critiques of voting methods on enacting reform we as a group make a point of making articles on wonky analysis abundantly clear that the system in question is still better or much better than both FPTP and IRV if that is the case. Since IRV is above and away the leading reform if we were to say, for example, "STAR fails FB" we would add in "...though it better meets this standard than either IRV or FPTP." 

In our world the question is not cardinal vs ordinal, but in the world at large that is absolutely the prevailing question for reform in any conversation we engage in. 

[Ciaran Dougherty]

> It is interesting that after all these years, you have yet to show a generally advisable STAR strategy that's better than honesty. One could call that a pretty strong endorsement of STAR Voting.

Um... isn't it obvious?

Score your favorite max (5), your least favorite minimum (0), and then count in from the edges from there.  If a candidate has a chance of beating your favorite, put them in the "count from the bottom" section, if they don't (or you're more concerned about the Worst candidate getting in), count from the top.

For example, if your honest preference were Bernie>Clinton>Johnson>Trump (actual degree of preference largely irrelevant), you would cast a B5,C4,J3,T0 or B5,C2,J1,T0, depending on whether you thought that Clinton could beat Bernie.

Under such a strategy, you would maximally help candidates you like get into the top two, minimize the chance that you help candidates that could beat them in the runoff would make the top two, and maintain your voice in the runoff.  How would that not be the optimal strategy? 

Heck, voters could even convince themselves that they're being relatively honest, because they're expressing relative preferences.  Indeed, they're being more at least as honest under such strategy as they are allowed to be under Ranked ballots...

[Ciaran Dougherty]

I like the idea of these, though I have some concerns about them.

 Obviously advocates for any system could cherry pick criteria they pass and elevate them, as FairVote has done,

 

OMG the cherrypicking.  On reddit's End FPTP forum, someone asked about FV's criteria, and a few people responded that the three they chose (LNH, Multi-Seat/Single-Seat versions, and Mutual Majority Criterion) appear to have been explicitly chosen to only support IRV/STV

 

1. Equality Criterion: The weight and worth of the citizens' votes is equal if the voter chooses to use the full weight of their vote. i.e. Any way that I fill out my ballot you can fill yours out in an equal and opposite way.

 

Excellent criterion.

2. The Wasted Vote Criterion: A voter cannot waste their vote by voting their conscience. ie. even if your favorite or favorites cannot win, your honest vote can always help you get better representation than if you had not voted at all. (An honest vote contains a voters honest preference order and degree of support.)

This one seems... odd to me.

 

3. Lesser-Evil Criterion: It is never a good strategy to rank or rate your lesser evil above your favorite. 

Only objection is the name.  Because that's No Favorite Betrayal, and/or Independence of Irrelevant Alternatives (which may be the same criterion, given that they have largely the same method of operation).

4. Consensus Criterion: If the ballots submitted in a given election contain information that shows that there is a candidate which has support from the entire electorate, the voting system must consider that information when selecting the winner that best represents the electorate. (This is the inverse of LNH)

With all due respect, the reason I am uncomfortable with STAR is that it doesn't satisfy this one (well, that, or the criterion itself is kind of meaningless).  Consider the scenario in this video, where one candidate gets 4 stars from everybody, but would lose under STAR. 

60% C5,S4,B1

40% C1,S4,B5

Average: C3.4, S4, B2.6

Under Score, S would win, with an average of 4.  Under STAR, C & S are the top two (S4.0>C3.4 || >B2.6), and C would beat S in the Runoff 60>40.

To my thinking, that means that either STAR doesn't meet this criterion (because it fails to elect the consensus candidate), or the criterion is meaningless ("Well, we considered that information, but then threw it away when actually deciding who won").   Sure, this criterion eliminates IRV from consideration (which is A Good Thing™), but... so does "Lesser-Evil"/No Favorite Betrayal/Independence of Irrelevant Alternatives.

5. Honest Incentive Criterion: Dishonest strategic voting, where a voter insincerely alters their preference order between candidates is not incentivized. ie. that strategy is more likely to backfire than it is to help them overall.

To me, this seems like a rephrasing of 3, except with broader scope, and qualifiers (order, not degree, of support is considered important).

 

I also would love to see the adoption of a walked back standard for election criteria as a whole. Pass/Fail is nice for mathematical proofs, but for instance there should be some sort of recognition if a system does a great job at both FB and LNH, while not getting 100% on either.

I would love that, too.  The trouble is how do you score them?  With absolute criteria there is a high degree of subjectivity as to whether a given method meets this criterion.  You might say, for example, that STAR gets a 95% on your Consensus Criterion, and I might say it's 87-93%, but... what would someone from FairVote say?

Would that be an Election Criteria Standard where a passing grade is 90% or 95%, but 100% is recognized as the gold standard? Can that kind of high but not perfect standard be objectively evaluated using simulated elections based on relative frequency of real world election types?

That is a legitimate and serious question... but the corollary question is "How do you define a whether a method meets it 90% of the time?"  After all, while we all (rightly) point to Burlington 2009, IRV worked exactly as advertised in San Francisco 2018.  Heck, for a majority of elections (where there is a clear and obvious winner), Plurality works fine.

Lastly, in light of what Jameson said above I'd like to propose a pact that in light of the real world implications of critiques of voting methods on enacting reform we as a group make a point of making articles on wonky analysis abundantly clear that the system in question is still better or much better than both FPTP and IRV if that is the case. Since IRV is above and away the leading reform if we were to say, for example, "STAR fails FB" we would add in "...though it better meets this standard than either IRV or FPTP." 

This seems sensible.  I actually texted a good friend of mine in Eugene a few weeks ago, urging her to support STAR (which she hadn't yet heard of [get on that, yo]), because it is, in my opinion the 2nd best possible single-seat method (after Score).

[Sara Wolf]

With Ciaran's honest strategy above I think he makes a good point. The key there is that "Indeed, they're being more at least as honest under such strategy as they are allowed to be under Ranked ballots..." 

I think it's important when we talk about expressive voting methods and strategy that we make a distinction between dishonest voting. Ie. Favorite betrayal, Lesser Evil Voting, or otherwise changing the order of your candidates, and other types of strategy where a voter may just chose to be less expressive or somewhat tactical, but to a limited extent. My theory is that dishonest voting is MUCH more disastrous to election accuracy overall. I'd love to see a VSE study that split those types of strategy into two categories and took a look at that. 

Countering Ciaran's honest strategy above I do still see some serious risk that would encourage a voter to be somewhat honest even when considering that strategy. 

For example, if your honest preference were Bernie>Clinton>Johnson>Trump (actual degree of preference largely irrelevant), you would cast a B5,C4,J3,T0 or B5,C2,J1,T0, depending on whether you thought that Clinton could beat Bernie.

Assuming that you see both Johnson as a non-threat and assuming you see Bernie as a shoe-in, sure that's a safe bet. But if there's a risk that Johnson or Trump could beat Bernie you have a strong incentive to score Hillary higher, if you would really be significantly happier with her. If a race is at all close the fully honest/expressive vote is still the best strategy. Thus, the more important your vote is to the election, the more the incentive to be honest and expressive. It's not perfect, but nothing is.

Compared to other systems, if this is the worst case strategic incentive I say we've done good. Note that the 0-5 scale was specifically selected to mitigate the impact and extent of this strategy. (and people like 0-5 WAY better!) 

People never want to vote lesser-evil again and for all intents and purposes with STAR they don't have to. 

[Ciaran Dougherty]

On Sun, Aug 19, 2018 at 1:44 PM, Sara Wolf <sa...@equal.vote> wrote:

Countering Ciaran's honest strategy above I do still see some serious risk that would encourage a voter to be somewhat honest even when considering that strategy. 

No, you saw a flaw in my example.  You seem to have ignored the strategy.  

 

Assuming that you see both Johnson as a non-threat and assuming you see Bernie as a shoe-in, sure that's a safe bet.

Yeah, that was presupposed given the "If a candidate has a chance of beating your favorite, put them in the "count from the bottom" section, if they don't (or you're more concerned about the Worst candidate getting in), count from the top." part of the strategy.

As such, if you considered Johnson a legitimate threat to Bernie, and Clinton not, sure, you could vote B5,C4,J1,T0.

Regardless, nothing you've suggested as a problem for the strategy is in conflict with the strategy as stated.  The principle stands.

If a race is at all close the fully honest/expressive vote is still the best strategy.

Again, the strategy holds in that case, too:  5 for your favorite and a 4 for whatever candidate you think could get into the top two and that your favorite could beat, and count upwards from zero for everyone else, maintaining candidate order. 

That way, if the top two is your Favorite (F), and your Backup (B), you will have maximally helped F beat B. 

If the top two are B and X, you will have maximally helped B beat X. 

If the top two are X and Y, you will have helped your favorite win.

Seriously, the only possible downside to this that I'm aware of is in a Condorcet Cycle, where you have to engage in either Favorite Betrayal, or Second-Favorite Betrayal.  ...which means that STAR also fails your Honest Incentive Criterion.

 if this is the worst case strategic incentive I say we've done good.

And I disagree because of why that is the worst case.   Yes, STAR voting gets rid of the "majority forcing their preference though min/max voting" strategy from Score, but it doesn't actually solve it.  It won't happen under STAR because the majority is going to get what they want even if that isn't the group consensus.  As such, that particular strategy becomes pointless; why spend effort trying to ensure that your candidate wins over a consensus candidate if the voting method will do that either way?

[Clay S]

On Saturday, August 18, 2018 at 11:11:48 AM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

But ok, the Gibbard-Satterthwaite theorem goes a long way toward proving that
honesty is not best strategy for STAR.  Which you should have already
known and hence never even raised this asinine "debate."

That theorem does not prove that you have discovered a HEURISTIC for casting a better-than-honest vote. Do you see the difference? 

[Sara Wolf]

I hear your points, Ciaran.
I agree with you that what you presented is the best "strategy," and that as you said, by normal standards that would be considered an "honest" strategy. I saw that too, and as I mentioned, that was the reason I fought so hard for the switch from 0-9 to 0-5. Beyond that there's not much else that can be done to mitigate the strategy without making other sacrifices. Gibbard-Statherwaite called it. The takeaway for me was that when I compare that strategic scenario for STAR, and I consider that well, at least it's not dishonest voting, and at least I believe (and VSE confirmed) that even if voters do that it's still quite accurate... We end up with a fair picture of what we have to work with there. It's not perfect, but it's pretty damn good. It's a lot better than the issues with IRV, which is the other leading option here in Oregon. And even other good systems have their own drawbacks to consider.

For comparison, let's look closer at strategy under Score Voting for the scenario from the video you cited above. I also like Squortle and would root for him in the video you linked, but if Score voters were strategic would Squortle have won? Considering that he was nobody's favorite it's likely that at least some people would have minimized him at least down to a 3 or bullet voted him to a zero to help their favorite. Some might have maximized him too. Hard to say. In any case I'm not 100% convinced that Squortle would have won under Score either. 

Strategic Score: 

60% C5,S3,B0

40% C0,S3,B5

Average: C3, S4, B2

Under Score, if voters minimized their 2nd choice 1 point only, to a 3, S would tie with C, with an average of 3. If even one voter bullet voted, Squortle would lose. If voters were 100% honest then I think plain Score Voting would be perfect, but considering the strategic incentives, I think that STAR is a great innovation. If you think that most voters won't be strategic, which may be true, I'd still think both are really good. I'd honestly love to see both passed and I will support the efforts of both Score and STAR activists.

I actually remember having this exact conversation in detail back when we were still choosing which reform to work on in Portland... at the time I was arguing for IRV or Score but no longer sold on either. At the end of it I wrote Mark and asked where I could order a shirt. Having looked honestly at the drawbacks of each option, including the fact that STAR was brand new (and still called SRV at that point,) I felt confident that working to develop STAR as an option was the right thing to do. 

 As you point out, Score is certainly less majoritarian than STAR, but STAR is less majoritarian than other options. Plus a lot of people would argue that Squortle wasn't preferred by the 60% so he didn't deserve to win. I'm not going to argue that. To be honest I see the above as one of the most comprehensive and fair critiques of STAR voting that I've ever seen. STAR is not perfect and I'd like to clarify, that while I love it and I hope we get it adopted all over the place, I don't think it's perfect. I wish it passed every criteria (except LNH), but I support it because I see it as a great option, as a consensus building option, and as the most well rounded proposal once we take everything into consideration. If you asked CES and FV and Counted to vote on a system: STAR Voting is Squortle. 

[Mark Frohnmayer]

Greetings election science fans. I'm sensing this thread has diverged a bit from the original topic, which concerns errors and misleading statements in RangeVoting's page regarding STAR Voting.

Why this is important?:

Despite it being clearly labeled in the first sentence as "PRELIMINARY" and Smith's admission in the second paragraph that "[Regarding STAR,] I'm biased against it and dubious," this page is being used by opponents of the system to regurgitate erroneous FUD on more widely-reachable sites like Wikipedia (see: https://en.wikipedia.org/wiki/Talk:STAR_voting#Motivation_section).

What statements on RangeVoting are in question?

Many sections of this page contains both objectionable content and some nuggets of wisdom. I'll start with a few of the former:

In "What Is STAR Voting?" Smith writes: "Perhaps a better name would have been STIR (score then instant runoff)."

Answer: Clay Shentrup wrote on 10/11/14 that STAR "need[s] the best possible name. I think 'rated' or 'rate' is bad because it's too liable to be confused with 'ranked' or 'rank'. 'Score' is a better word, in general. But 'scored' is not a great word either: Scored Instant Runoff Voting = hrmmmmm[.] Crack that nut."

We tried Rated Instant Runoff Voting (R-IRV). We tried Score Runoff Voting (You got SRV'd). Everyone got confused... "Oh is this rank score voting?"

The name "STAR Voting" immediately connects voters with the ubiquitous star rating systems that are used to rate apps and products, which makes it easy to draw distinction between it and RCV/IRV. STIR might play in the Reggae community ("STIR IT UP!"), but for the world at large, STAR is a good name. To be clear, I fought STAR Voting for a long time, but Sara Wolf is quite persuasive when obviousness is on her side.

In "Initial comparison of STAR versus plain range and versus R+R" Smith opines that R+R will have better results than STAR with strategic voters.

Answer: Smith's analysis rests on the honesty of the second (runoff) vote, but fails to consider what could happen when you can be dishonest in the first step and honest in the second. Specifically, the bogus "burying" concern FairVote raises regarding STAR (see http://equal.vote/fv) is a legitimate concern in R+R, precisely because the voter gets the opportunity to correct the record. It is not clear that Smith has either considered or simulated this significant difference between STAR and R+R. In this section it appears Smith misquotes his own work - see https://rangevoting.org/StratHonMix.html where in a 50/50 strategic to honest vote R+R beats Range (not 75%). Of course, none of Smith's hypotheses about what his own simulations would show using STAR are worth anything, as he has not actually run such simulations. Actual data, which we've asked for since 2014, would be useful here.

In "Properties sacrificed by STAR voting," Smith lists "monotonicity" as the first of the "nice theoretical properties" that STAR throws in the garbage.

Answer: Misleading. As Jameson Quinn wrote above in this thread, this is "an attempt to redefine monotonicity to something you think is more important. You may or may not be right, but I still think that "monotonicity" should not be the word for your new concept. STAR is monotonic." Smith clarifies later that he is using monotonicity in a "slightly wider sense of the word" which has lead directly to Markus Schulze to claim on Wikipedia that STAR violates the Monotonicity Criterion. If Smith wants to introduce a NEW criterion that includes his "slightly wider sense of the word," he should. I've got an acronym or two that might stick. But to use the well-defined term "monotonicity" to diss STAR here is incorrect and actually damaging.

In "Voter Honesty," Smith asserts that "Voters are motivated to lie in their STAR votes," plays the https://en.wikipedia.org/wiki/Reductio_ad_Hitlerum fallacy, and suggests that voters may find it best to give their favorites a zero

Answer: As Clay Shentrup notes above, Smith fails entirely to provide anything resembling a real heuristic voters can use to determine when they should adopt Smith's proposed strategy before the election occurs. We believe a more realistic assessment of strategic voting options with STAR is presented at https://www.equal.vote/strategic-star.

I could go on...

In closing:

...but it's late. Please keep the discussion on this thread on the stated topic of RangeVoting's STAR analysis. Other strategic implications of STAR, campaign efforts, etc. can be discussed elsewhere, but I'd really like to see something approaching an unbiased consideration of STAR on RangeVoting ASAP. Any help you all can offer in this regard would be much appreciated.

Cheers,

Mark

[Mark Frohnmayer]

https://en.wikipedia.org/w/index.php?title=STAR_voting&action=history

Some late night fun.