OpaVote post favoring Approval Voting to IRV

[Clay Shentrup]

Someone at OpaVote wrote this incredibly uninformed opinion favoring IRV to Approval Voting, based on obvious logical fallacies.

http://blog.opavote.com/2016/10/why-i-prefer-ranked-choice-voting-to.html#comment-form

My response:

You're making common logical fallacies.

First, it doesn't make sense to say that you prefer voting system X to voting system Y. You prefer _candidate_ X to _candidate_ Y. Thus the measure of a voting system is simply how happy you are with the winners, on average. This can be measured via on objective measure called Bayesian Regret. Approval Voting does extremely well, while IRV is the worst of the five commonly discussed alternatives.

http://ScoreVoting.net/BayRegsFig.html

> has the least cognitive burden among the three methods discussed here

This is drastically wrong. You can see the multitude of ways that IRV is objectively more complex than systems like Score Voting and Approval Voting here.

http://scorevoting.net/Lorenzo.html

IRV (or any ranked system) has a higher rate of spoiled ballots, so voters empirically mess up their ballots more often than with Score Voting, Approval Voting, or even Plurality Voting.

http://scorevoting.net/Lorenzo.html

Cardinal methods are computationally simpler to complete because they only require two passes. One to find the min and max utility, and the second to normalize and output them. Whereas ranking is O(n^2), since a human brain uses bubble sort not quicksort.

IRV is also one of the worst systems in terms of intuitive understanding.

https://sites.google.com/a/electology.org/www/approval-score-sf#TOC-Intuitive-understanding

> This kind of ballot has low cognitive burden because a voter doesn't have to consider which candidates are viable.

Simply false. IRV fails the Favorite Betrayal Criterion, so if your favorite candidate isn't "electable", you shouldn't rank her in first place. E.g. some Republicans in the 2009 IRV mayoral race in Burlington, VT could have gotten their second choice instead of their third if they had insincerely ranked their second in first place. Kind of like voting Democrat when you really prefer Green, so you don't get Republican.

http://scorevoting.net/Burlington.html

> Voting geeks cause this the later-no-harm criterion.

Later-no-harm is an "anti-criterion". Satisfying it is BAD. Explanation by a Princeton math PhD and top voting theory expert here.

http://scorevoting.net/LNH.html

A key takeaway there is that with IRV, ranking an additional candidate CAN hurt YOU the voter.

>  The possibility, however, that your second and later choices hurt your first choice is so small that, for practical purposes, a voter to cannot take this into account

Ludicrous. It's not absolute probabilities that matter, but _relative_ probabilities. A Green voter (for example) has to ask herself which is more likely to occur if she insincerely ranks the Democrat in first place:

1) That will cause the Democrat to win instead of the Republican (GOOD!)

2) That will cause the Democrat to win instead of the Green (BAD!)

Since #1 is obviously more likely than #2, the voter is advised to insincerely vote for the major party candidate (or candidate most perceived to be electable) 100% of the time. Further mathematical explanation here.

http://scorevoting.net/TarrIrv.html

This isn't even getting into practical considerations like precinct summability, transparency of results, propensity to result in duopoly, etc. I cover some of those other factors in a layman friendly video here.

https://www.youtube.com/watch?v=2q_eMUGCU5U&t=487

IRV is simply the worst of the five commonly discussed alternative methods by any informed assessment.

Clay Shentrup

Co-founder, The Center for Election Science

[Brian Kelly]

Good evening Clay,

Nice job working to set the record straight.  You make a lot of good points and I hope Jeff will dig into them.

I did have one thought I wanted to share.

> Later-no-harm is an "anti-criterion". Satisfying it is BAD.

I understand why you say this but I think it's a matter of perspective.  It comes back to my question "What is a vote?"

Many people see a vote like a voucher that they give to their favored candidate and they only want it passed to their second choice where there is no chance the first will win.  This is LNH and these kinds of people prefer IRV.

Other people see voting as a rating of a candidate and see that someone's second or third choice is indicative of value which a system can use.  This view works very well with score voting systems and is incompatible with LNH.  You are obviously in this camp.

So LNH is actually a proxy for a deeper difference of opinion so I think you will make little headway attacking it.  It might be better to ask if the person thinks an election is about finding the person the most people like the best or finding the person the people like the most.

[Clay Shentrup]

On Monday, October 10, 2016 at 10:32:15 PM UTC-7, Brian Kelly wrote:

> Later-no-harm is an "anti-criterion". Satisfying it is BAD.

I understand why you say this but I think it's a matter of perspective.  It comes back to my question "What is a vote?"

It's not a matter of perspective. Ignoring intensity of preference is literally a form of information loss. 

[Toby Pereira]

On Tuesday, 11 October 2016 04:32:19 UTC+1, Clay Shentrup wrote:

First, it doesn't make sense to say that you prefer voting system X to voting system Y. You prefer _candidate_ X to _candidate_ Y.

Sometimes I think you're difficult for the sake of it. You yourself prefer approval voting to IRV because you think it's a better system. It makes perfect sense.

[Steve Cobb]

I think the argument for LNH is that it reduces the incentive for tactical voting. This is true, but only for one form of tactical voting, the least perverse. IRV advocates seem willing to accept a direct reduction in fairness in exchange for a possible small reduction in tactical voting.

[Clay Shentrup]

Toby,

No. Approval has lower Bayesian Regret than IRV so it elects *candidates* that I prefer.

[Jameson Quinn]

Clay:

I understand that you think that BR (aka VSE) is just about the only measure of an election system that matters. I understand your reasons for taking that position, and I agree with most of them.

But please understand: Warren's BR numbers are not the be-all and end-all of utilitarian comparisons of voting systems. They were a very good start, and there are certain things that they do show pretty conclusively. But they do rely on certain assumptions, some of them questionable. It's much better to say "here's some strong evidence that X is true and Y is false" rather than "X is true, Y is false, it is known, and anyone who doesn't agree is stupid".

2016-10-12 11:11 GMT-04:00 Clay Shentrup <cl...@electology.org>:

Toby,

No. Approval has lower Bayesian Regret than IRV so it elects *candidates* that I prefer.

--
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.

[Clay Shentrup]

Jameson,

The accuracy of Warren's particular BR calculations is beside the point. The point is, again, that you prefer candidates not voting methods. I general unqualified statement that "I prefer IRV to Approval Voting" is indicative of general ignorance of social choice theory.

And even to the extent Warren's figures are imperfect, they are still the best I'm aware of by far.

[Brian Kelly]

Clay, your argument is circular.

You believe votes are ratings, not vouchers.

So...

Score voting is the best solution.

Because...

Doing otherwise is a loss of valuable information.

Because...

Votes are ratings, not vouchers.

[Clay Shentrup]

On Friday, October 14, 2016 at 9:00:40 AM UTC-7, Brian Kelly wrote:

Clay, your argument is circular.

No, it isn't.

You believe votes are ratings, not vouchers.

I have no idea what a "voucher" is supposed to mean. Preferences are cardinal utility values. See mathematical proof here.

http://scorevoting.net/UtilFoundns.html

[Clay Shentrup]

Also, the fact that preference information includes intensity is not remotely circular. It's just empirically demonstrable.

http://scorevoting.net/Mill

[Brian Kelly]

I have no idea what a "voucher" is supposed to mean.

So you are saying that I am wrong when you don't even grasp what I am saying?

Wow.

Clay, I am willing to put in the time to follow your logic and proofs but I need to know you will put in the time to understand what I am saying if/when I have a concern.  Tell me, is that going to happen?

[Clay Shentrup]

On Friday, October 14, 2016 at 9:29:58 AM UTC-7, Brian Kelly wrote:

I have no idea what a "voucher" is supposed to mean.

So you are saying that I am wrong when you don't even grasp what I am saying?

I don't think what I just wrote was at all unclear. I said that:

1) you're wrong about my argument being circular (this is a common fallacy made by people who don't seem to understand that scores on Score Voting ballots aren't utilities).

2) There is a mathematical proof that the social welfare function is just the sum of individuals' personal cardinal utilities.

Clay, I am willing to put in the time to follow your logic and proofs but I need to know you will put in the time to understand what I am saying if/when I have a concern.  Tell me, is that going to happen?

Look, you used a vague unspecified term ("vouchers") that has no existing meaning in the field of social choice theory. If you want people to know what you're talking about, just explain what you're talking about. 

[Brian Kelly]

Look, you used a vague unspecified term ("vouchers") that has no existing meaning in the field of social choice theory. If you want people to know what you're talking about, just explain what you're talking about.

I defined "voucher" in my first post in this thread.  Did my description there make sense of how people might see a vote as either a voucher or a rating?

[Neal McBurnett]

I wish you two wouldn't waste your time and our time with the tendency both you have to leave out important context when responding.

Re "vouchers", Brian wrote:

> I understand why you say this but I think it's a matter of perspective.  It comes back to my question "What is a vote?"
>

> Many people see a vote like a voucher that they give to their favored candidate and they only want it passed to their second choice where there is no chance the first will win.  This is LNH and these kinds of people prefer IRV.
>
> Other people see voting as a rating of a candidate and see that someone's second or third choice is indicative of value which a system can use.  This view works very well with score voting systems and is incompatible with LNH.  You are obviously in this camp. So LNH is actually a proxy for a deeper difference of opinion so I think you will make little headway attacking it.  It might be better to ask if the person thinks an election is about finding the person the most people like the best or finding the person the people like the most.

Brian, I don't think it makes sense in general to consider a vote to be a "voucher". What do you do when a voter has equal preference for two candidates? Who do they give their voucher to? Does the candidate get to keep the voucher and decide what is done with it next (as in some of the simple, efficient Canadian voting proposals discussed earlier), or is the voter required to specify rankings as well as ratings?

And Clay, of course people have preferences with respect to voting systems. You're welcome to continue to argue that there is actually good science which they should base their preferences on, and of course I agree in many cases. But as pointed out earlier, we really don't have any iron-clad scientific arguments that any one particular voting system is the best, unless we have unprecedented information about the preferences of a population for not just candidates, but voting methods.

And yes, people in a democracy get to decide on a voting system based on their preferences. Voting systems aren't decreed by experts who consult their favorite theories and data.

Neal McBurnett http://neal.mcburnett.org/

> --
> 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.

[Brian Kelly]

> I wish you two wouldn't waste your time and our time with the tendency both you have to leave out important context when responding. 

Sorry Neal.  I am trying to be respectful of my audience's time but perhaps too much so.  I will cut down on the cutting.

> Brian, I don't think it makes sense in general to consider a vote to be a "voucher". What do you do when a voter has equal preference for two candidates?  Who do they give their voucher to?

It's a good question, Neal.  Along those same lines, since many in the US go to the polls to vote against what they perceive as the greater evil, wouldn't an "anti-voucher" (a voucher to cancel someone else's voucher) be more fitting to the actual goals of many voters?

My original point to Clay was that I think asking these sorts of questions serves the discussion much better than saying, "Later-no-harm is an 'anti-criterion'. Satisfying it is BAD."  Not only does it encourage critical thinking but I believe it brings to the foreground that actual reasons for wanting later-no-harm.

[Toby Pereira]

On Wednesday, 12 October 2016 16:11:44 UTC+1, Clay Shentrup wrote:

Toby,

No. Approval has lower Bayesian Regret than IRV so it elects *candidates* that I prefer.

And it therefore makes sense to say you prefer it as a method. But this seems to be nothing more than word games.

[Toby Pereira]

But also if there was a vote on what voting system we should use, then the voting systems would be the candidates. So would it be OK to say which voting system you prefer then?

But equally, you could say that you don't prefer candidates but the legislation they propose/vote for. Or that it's not even the legislation but the effects of that legislation. 

[Clay Shentrup]

On Saturday, October 15, 2016 at 2:21:09 PM UTC-7, Brian Kelly wrote:

I defined "voucher" in my first post in this thread.  Did my description there make sense of how people might see a vote as either a voucher or a rating?

I would call that a partial description of "voucher" not a definition. But either way, it has problems:

> Many people see a vote like a voucher that they give to their favored candidate and they only want it passed to their second choice where there is no chance the first will win.  This is LNH and these kinds of people prefer IRV.

Is this even a meaningful statement? I don't think so.

"There is no chance the first will win." -- How should we determine who has a chance to win in the first place? That's the very point in contention. If your "voucher" concept says that none of my support should go to anyone until she "can't win", then the only measure of support we have is first-place rankings. So "voucher" is just a substitute for "IRV ballot".

A vote is a statement of a voter's opinion. An election the aggregate of all voters' opinions.

>  This is LNH and these kinds of people prefer IRV.

Well, they think they prefer IRV, because they're confused. What they typically don't realize is that LNH assumes you want to start by ranking your favorite in first place. In many cases you don't, so LNH is irrelevant.

Moreover, LNH is about hurting candidates, whereas tactical voting is about hurting voters. IRV can allow you to hurt yourself by ranking more candidates.

[Clay Shentrup]

On Saturday, October 15, 2016 at 3:01:03 PM UTC-7, Neal McBurnett wrote:

And Clay, of course people have preferences with respect to voting systems.

That doesn't make sense. The output of a voting system is an election winner. That is the thing you have preferences about. 

[Clay Shentrup]

Well, no. I prefer the candidates it elects. Even if you say I prefer the voting method itself, at least that's ultimately based on the result utility, not some absurd gut feeling I have about one system or another. 

[Warren D Smith]

I don't understand what you all are arguing about.


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

[Steve Cobb]

We recently had a discussion about the three kinds of voting methods: allocational, ordinal, and cardinal. Allocational methods are the traditional kind, where each voter gets a fixed amount of vote that he or she may divide among candidates. The allocation may be continuously divisible (e.g. 1/2, 1/3, and 1/6 to your top three candidates) subject to constraints like discrete tokens (e.g. cumulative voting). Jameson's EPH method is allocational with the constraint that each chosen candidate receives an equal 1/N amount.

Note that "ballot" originally meant what you are now calling a voucher and what I am calling a token:

https://en.wikipedia.org/wiki/Ballot#Etymology

[Clay Shentrup]

On Sunday, October 16, 2016 at 8:43:29 AM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

I don't understand what you all are arguing about.

I'm mainly getting at the fallacy inherent to statements like "I prefer IRV to Approval Voting because.."

This is a fallacy because its focusing on some subjective property instead of Bayesian Regret. Or as I often say, you don't prefer voting methods, you prefer candidates. The measure of a voting method is just how much you typically like the resulting candidates it elects. 

[Steve Cobb]

I can't say I prefer Approval Voting to IRV because the latter requires too much effort to count?

http://wiki.electorama.com/wiki/Summability_criterion