STLR - Score Than Leveled Runoff might not be too complex for voters

[Lime]

@toby-pereira said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Using score ratios is always a bad idea. It only makes sense (as with utility scores) to look at the absolute difference between scores. If someone scores three candidates 0, 1, 2, then they are equidistant. To see the 1 as infinitely more than the 0 is bad voting method behaviour.

It makes more sense under this model of voting behavior:

1. The utility of a candidate for Congress is defined to be equal to the (importance-weighted) probability that they will cast a tiebreaking vote agreeing with your preferences. (Times the number of categories).
2. This probability is never actually 0; it's bounded between 0 and 1.
3. Voters normalize their ballots by dividing by the probability for the worst candidate on the ballot.

The reason : I don't see that much evidence that honest voters normalize their ballots. Strategic voters tend to more-than-normalize it.

Empirically, people are quite good at performing interpersonal comparisons of utility. Their guesses are very rough at the individual level, but in the aggregate, this cancels out. Usually there's a substantial proportion of unnormalized ballots (which lower-bounds the probability of strategy).

On the other hand, strategic voters will probably more than normalize their ballot (most will be min-maxing).

@gregw said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Would Star be a good choice for a single-winner election with no more than 5 candidates, several of the candidates could be from the same party? Would similar candidates cause a clone problem? (The candidates would not be able to form alliances.)

Pretty much no voting system other than FPP has cloning problems. The only potential problem for STAR is a lack of clones, which can could in theory, very rarely, cause the strategic incentives to go a bit haywire. The more candidates running, and the more ideologically similar candidates, the better.

[GregW]

@toby-pereira said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Using score ratios is always a bad idea. It only makes sense (as with utility scores) to look at the absolute difference between scores. If someone scores three candidates 0, 1, 2, then they are equidistant. To see the 1 as infinitely more than the 0 is bad voting method behaviour.
It would also be insane if (under a 0 to 5 ballot) if someone scoring two candidates 5 and 1 had less voting power in the run-off than someone scoring them 1 and 0.

Therefore, STLR’s leveled runoff is not a good idea, but it might be less bad using a 1 - 6 scoring range than a 0 - 5 range. Also, normalizing a ballot with scores of 4, 2, 1, and 0 to scores of 5, 3, 2, and 1, keeping the absolute differences, makes more sense. Call it STLR2? It would be a little easier to explain the regular STLR. Thank you for the help!

[GregW]

@lime said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Pretty much no voting system other than FPP has cloning problems. The only potential problem for STAR is a lack of clones, which can could in theory, very rarely, cause the strategic incentives to go a bit haywire. The more candidates running, and the more ideologically similar candidates, the better.

Therefore, FPP is the worst possible method for a single-winner party primary race with many candidates. STAR would be a better choice. Thank you for your help!

[Toby Pereira]

@gregw said in STLR - Score Than Leveled Runoff might not be too complex for voters:

@toby-pereira said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Using score ratios is always a bad idea. It only makes sense (as with utility scores) to look at the absolute difference between scores. If someone scores three candidates 0, 1, 2, then they are equidistant. To see the 1 as infinitely more than the 0 is bad voting method behaviour.
It would also be insane if (under a 0 to 5 ballot) if someone scoring two candidates 5 and 1 had less voting power in the run-off than someone scoring them 1 and 0.

Therefore, STLR’s leveled runoff is not a good idea, but it might be less bad using a 1 - 6 scoring range than a 0 - 5 range. Also, normalizing a ballot with scores of 4, 2, 1, and 0 to scores of 5, 3, 2, and 1, keeping the absolute differences, makes more sense. Call it STLR2? It would be a little easier to explain the regular STLR. Thank you for the help!

Well, if I was normalising, I would always stretch scores to fill the whole range. So every voter would have a 0 and a 5 in this case.

[Jack Waugh]

STLR and STAR are both slight elaborations on top of plain Score Voting. My opinion is that for a single winner, the best of the three, in terms of the power relations created, is plain Score. As I understand it, the main critique of plain Score by the advocates of the elaborations on top of it is that plain Score gives more incentive than the elaborations do to vote "tactically" meaning taking into account a guess or estimate of what other voters think, rather than simply reflecting ones own preferences. I counter that criticism by arguing that tactical voting is not harmful in the context of systems that conform to the balance constraint and are additive (both of which are true of Score and STAR and I guess STLR) provided that all parties know how to do it.

Now when any individual of us or any group consisting of some of us is thinking about what to push hard for with the public, we have grounds to think not just about what systems would be adequate for the purpose, but also, from among the reasonable systems, about which of them we are likely going to succeed at selling. @SaraWolk has shown in this forum one or more examples where the STAR advocacy group has approached a maker of voting machines, saying, "why don't you implement STAR", and the reply has come back, "we're waiting for the experts to finish bickering and settle on something".

For a runoff for a single winner, what will be easiest to sell? There are two characters in my head arguing opposite sides of this question. Here's how these arguments go:

• Plain Score may easier to sell, because it is simpler by half. There is just one round of tallying and just one type of round of tallying, as opposed to two rounds of tallying and two types of round of tallying for STAR.

• STAR may be easier to sell, because for purposes of selling it, one can refer to all the propaganda (not necessarily false or misleading) available from the STAR advocacy group, which is quite persuasive because of all the points it gives in favor, none of which is necessarily horribly incorrect, and the rich history that it includes, which includes that there was inspiration from goals or subgoals that had been stated by the "RCV" advocacy group or thinkers close to them.

[GregW]

@toby-pereira said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Well, if I was normalising, I would always stretch scores to fill the whole range. So every voter would have a 0 and a 5 in this case.

That would be fair and accurate. Might be a little more difficult to explain.

[GregW]

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:
(regarding Score)

tactical voting is not harmful in the context of systems that conform to the balance constraint and are additive (both of which are true of Score and STAR and I guess STLR) provided that all parties know how to do it.

My concern with Score is more with habits than tactics. Radicals might habitually be inclined to give their favorites the top score and everybody else the lowest score, while moderates may usually be more nuanced, giving radicals, over time, an advantage.

Perhaps I too much of a worry wort. Are the current instructions for Score voters:
"Score the best candidate 5 and the worst candidate 0. Score the other candidates in comparison to the best and worst candidates. Ties are allowed."
adequate protection against tactics and habits?

Jack Waugh also stated:

Plain Score may easier to sell, because it is simpler by half. There is just one round of tallying and just one type of round of tallying, as opposed to two rounds of tallying and two types of round of tallying for STAR.

STAR may be easier to sell, because for purposes of selling it, one can refer to all the propaganda (not necessarily false or misleading) available from the STAR advocacy group, which is quite persuasive because of all the points it gives in favor, none of which is necessarily horribly incorrect, and the rich history that it includes, which includes that there was inspiration from goals or subgoals that had been stated by the "RCV" advocacy group or thinkers close to them.

For those reasons, Score would be easier to sell on a sidewalk or a stranger’s front porch. STAR may be easier in print.

[Lime]

@gregw said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Therefore, FPP is the worst possible method for a single-winner party primary race with many candidates. STAR would be a better choice. Thank you for your help!

Wait wait wait, that's not necessarily true

There's a big problem with holding any kind of party primary at all. You can find a very good post about this here. The issue is that candidates that are very representative of their party's beliefs can be very bad representatives for the beliefs of the electorate as a whole.

In this situation, the quality of different methods might even reverse: methods that are very bad, almost as bad as selecting a winner "at random", can end up being better. (Because this gives candidates who are "extreme" within their own party—i.e. very moderate—a shot at winning thanks to blind luck.)

[Lime]

@gregw said in STLR - Score Than Leveled Runoff might not be too complex for voters:

My concern with Score is more with habits than tactics. Radicals might habitually be inclined to give their favorites the top score and everybody else the lowest score, while moderates may usually be more nuanced, giving radicals, over time, an advantage.

This would give radicals less influence over the election overall: if they're radicals, their "favorites" will be too radical to win, and they'll lose all their influence by not rating any more moderate (electable) candidates.

[GregW]

@lime said in STLR - Score Than Leveled Runoff might not be too complex for voters:

There's a big problem with holding any kind of party primary at all. You can find a very good post about this here. The issue is that candidates that are very representative of their party's beliefs can be very bad representatives for the beliefs of the electorate as a whole.

I am very aware of primary problems, here is an excerpt from an article about Top Four and Fina Five voting system I am writing for my upcoming website VoterTakeCharge.US :

Taxpayer funded primaries give the Democratic and Republican parties a ton of free publicity as their candidates glide into the general election on the taxpayer’s dime. Third parties and unaffiliated candidates are required to jump through hoops to get on the ballot.

Worse, in safe districts, small sets of voters, often less than 10% of all registered voters, decide elections in the dominant party’s primaries. These voters tend to be more partisan than most voters. In contested races, two small sets of partisan voters nominate the only candidates that have a fair chance of winning.

Perhaps, more of the blame should go to the voters who do not show up. Many of them have a good excuse. They are not allowed to show up.

The lack of proportional representation, our current partisan primary system, and plurality voting are the pillars of two-party rule in America.

Changing from plurality voting to a better single-winner voting method may put a modest dent in the two-party system. Changing from major party primaries to nonpartisan primaries may cause significant damage to the two-party system. Proportional representation can break the two-party duopoly and more accurately voice the will of the voters.

[GregW]

@lime said in STLR - Score Than Leveled Runoff might not be too complex for voters:
Regarding bullet voting in Score elections

This would give radicals less influence over the election overall: if they're radicals, their "favorites" will be too radical to win, and they'll lose all their influence by not rating any more moderate (electable) candidates.

This is a very good argument, I may borrow it in the future.

[Jack Waugh]

@gregw said in STLR - Score Than Leveled Runoff might not be too complex for voters:

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:
(regarding Score)

tactical voting is not harmful in the context of systems that conform to the balance constraint and are additive (both of which are true of Score and STAR and I guess STLR) provided that all parties know how to do it.

My concern with Score is more with habits than tactics. Radicals might habitually be inclined to give their favorites the top score and everybody else the lowest score,

If that is how they want to vote, they have that right.

while moderates may usually be more nuanced, giving radicals, over time, an advantage.

Irrespective of which people you mean when you say "moderates" and "radicals", what kind of vote do you think is "more nuanced"? How does it give their opponents an advantage? Score is "one person, one vote", and those votes have equal influence. Only by making a blunder, an error of not following the optimal tactic, can a voter concede power to an opposing voter.

I don't know that voters have more than three categories in which they place candidates:

• love
• could accept
• Adolph Hitler

The right way to vote is to give full scores to the ones you love, and if you think other voters don't agree, give .99 of the full score (assuming the lowest score is zero) to the ones you could accept, and the lowest score to everyone else.

Perhaps I too much of a worry wort. Are the current instructions for Score voters:

There shouldn't be any instructions other than:

• explaining their freedom of movement in filling out the ballot without spoiling it, and
• explaining how the tally will work.

...
For those reasons, Score would be easier to sell on a sidewalk or a stranger’s front porch. STAR may be easier in print.

I suspect that "front-porch" (and elevator) selling is more important than in print. Americans have short attention spans. So I'm in favor of our pushing Score Voting.

If there are 99 or 100 or 101 possible scores, computers will be needed. If there are five or seven, the ballots can be tallied by hand. But as I mention above, I think that .99 of the way between the min and the max is tactically useful for supporting second-choice candidates when the first-choice ones are probably unpopular. So, I think the allowable scores should include 100, 99, and 0. For formal balance, 1 should also be allowed. All in all, I think the range {100, 99, 90, 50, 10, 1, 0} would be good. Leaving out 50 wouldn't worsen it significantly.

[GregW]

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Irrespective of which people you mean when you say "moderates" and "radicals", what kind of vote do you think is "more nuanced"?

To me radicalism is a “my way or the highway” state of mind rather than a preference for any political philosophy or public policy. A philosophy or policy can be radical or moderate, depending on the current most popular opinion in a society. A more nuanced voter would give more consideration to alternatives than “my way or the highway”.

If there are 99 or 100 or 101 possible scores, computers will be needed. If there are five or seven, the ballots can be tallied by hand. But as I mention above, I think that .99 of the way between the min and the max is tactically useful for supporting second-choice candidates when the first-choice ones are probably unpopular. So, I think the allowable scores should include 100, 99, and 0. For formal balance, 1 should also be allowed. All in all, I think the range {100, 99, 90, 50, 10, 1, 0} would be good. Leaving out 50 wouldn't worsen it significantly.

If we use Score with a {100, 99, 90, 50, 10, 1, 0} range, would the instructions be:
“Give each candidate a score of 100, 99, 90, 50, 10, 1, or 0.
100 is the highest possible score, 0 is the lowest.
Ties are allowed.
Blanks are equal to 0
The candidate with the highest total score wins."

Would you also add something like:
“A 99 score is tactically useful for supporting your second-choice candidate when the first-choice candidate might not be popular enough to win.”

[Jack Waugh]

@gregw said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Would you also add something like:
“A 99 score is tactically useful for supporting your second-choice candidate when the first-choice candidate might not be popular enough to win.”

No, I wouldn't include that in official instructions or information. It is only a personal opinion. The other parts you said seem right.

Please give an example of a nuanced vote and explain how you think it could work to the disadvantage of the person or faction that casts it.

[GregW]

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

Please give an example of a nuanced vote and explain how you think it could work to the disadvantage of the person or faction that casts it.

A Score voting scenario:
Faction A gives a score of 4 to a candidate they think is “pretty good” because of the imperfections of their candidate.
Faction B gives a score of 5 to a candidate they think is “pretty good” because they don’t give a damn about imperfections.
The Faction B candidate gets more score per ballot.
Thus the interest in normalizations and/or this Score instruction:
“Score the best candidate 5.
Score the worst candidate 0.
Score the other candidates in comparison to the best and worst candidates.
Ties are allowed.”

[Jack Waugh]

@gregw, then I predict that Faction A will within a few elections figure out that it should use the maximum value (5) and minimum value (0) of the permitted range. I think factions don't usually voluntarily give up power. Elections are contentious.

[Lime]

@gregw said in STLR - Score Than Leveled Runoff might not be too complex for voters:

“A 99 score is tactically useful for supporting your second-choice candidate when the first-choice candidate might not be popular enough to win.”

I wouldn't add anything about tactics; much better to avoid discussing it. I'd rather encourage voters to give honest ratings of each of the candidates, so we can get rid of spoiler effects. Instructing them on how to normalize ballots increases the risk of a spoiled election.

[Jack Waugh]

@lime said in STLR - Score Than Leveled Runoff might not be too complex for voters:

...

I wouldn't add anything about tactics; much better to avoid discussing it. I'd rather encourage voters to give honest ratings of each of the candidates, so we can get rid of spoiler effects; instructing them on how to vote tactically (or worse still, instructing them to normalize ballots) increases the rate of spoiled elections.

I disagree.

First off, let's separate official communication about an election from communication from a person's or a group's political takes.

The official communication about an election should indeed avoid laying out or suggesting tactics. It should only state the freedom of movement the voter has in filling out the ballot without invalidating it, and how the tally will work to determine the winner.

In non-official channels of communication, I see urging "honesty" as problematic and dishonest. Voting is not an opinion poll; it is an exercise of political power. It's like steering a boat. When you command "right full rudder", it's not an opinion, but a muscular exercise that feeds into the whole dynamic of the boat's motion in accord with the Laws O' Physics (TM), the design of the boat, the propeller's rotational velocity, etc.

What reasoning leads you to think that Score voters pushing exaggerated support hose toward compromise candidates tends to spoil elections? They are still giving more support to their true favorites. If enough proportion of voters are standing with them, that candidate can win.

Urging "honest" votes as though the election were an opinion poll is just sucker bait. People who follow your urging are giving up power to their opponents.

The Gibbard theorem showed that optimal voting takes any guesses or estimates of the positions of other voters into account.

[Lime]

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

The Gibbard theorem showed that optimal voting takes any guesses or estimates of the positions of other voters into account.

I'm not sure what you're talking about here. Gibbard's theorem just proves there's no single "best" strategy for voting in an election. The socially-optimal outcome is only possible if every voter is fully honest.

Enforcing automatic strategy only improves outcomes if either we think only one side will behave strategically.

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

In non-official channels of communication, I see urging "honesty" as problematic and dishonest. Voting is not an opinion poll; it is an exercise of political power. It's like steering a boat. When you command "right full rudder", it's not an opinion, but a muscular exercise that feeds into the whole dynamic of the boat's motion in accord with the Laws O' Physics (TM), the design of the boat, the propeller's rotational velocity, etc.

Or we could provide true information and help the system pick a better winner. If we're not going to let voters be honest (as many of them choose to be), what's the point in anything but a maximal lottery? If voters provide true, accurate information, the outcomes are better; when people vote honestly in score, the outcome is the utilitarian winner.

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

then I predict that Faction A will within a few elections figure out that it should use the maximum value (5) and minimum value (0) of the permitted range. I think factions don't usually voluntarily give up power. Elections are contentious.

Empirically, about 60% of voters choose to do so. When asked if they'd prefer to have a voting strategy automatically executed for them, 57% refused. When asked whether they thought voting systems should automatically execute strategic voting, opponents outnumbered supporters 2:1.
https://www.proquest.com/docview/304273753?pq-origsite=gscholar&fromopenview=true&sourcetype=Dissertations & Theses

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

What reasoning leads you to think that Score voters pushing exaggerated support hose toward compromise candidates tends to spoil elections? They are still giving more support to their true favorites. If enough proportion of voters are standing with them, that candidate can win.

Arrow's theorem, which implies that if voters base their on strategic considerations, there will always be spoiler effects. (Compare honest score voting, which is completely spoilerproof.)

[Jack Waugh]

@lime said in STLR - Score Than Leveled Runoff might not be too complex for voters:

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

The Gibbard theorem showed that optimal voting takes any guesses or estimates of the positions of other voters into account.

I'm not sure what you're talking about here. Gibbard's theorem just proves there's no single "best" strategy for voting in an election.

Let's read from the introductory part of the Wikipedia article on Gibbard's theorem:

for any deterministic process of collective decision, at least one of the following three properties must hold:

1. The process is dictatorial, i.e. there is a single voter whose vote chooses the outcome.
2. The process limits the possible outcomes to two options only.
3. The process is not straightforward; the optimal ballot for a voter depends on their beliefs about other voters' ballots.

I think you will agree with me that condition 1 does not apply to Score Voting if there is more than one voter, and that elections are possible in which condition 2 does not hold, either. That leaves us with a certainty that condition 3 holds. The optimal vote does not depend merely on the desire of the voter toward the candidates. It must also take into consideration whatever partial knowledge or probability estimates the voter feels in regard to the other voters.

The socially-optimal outcome is only possible if every voter is fully honest.

What grounds do you have for coming to such an opinion? I don't think it is correct.

Enforcing automatic strategy...

For what purpose do you introduce such a topic?

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

In non-official channels of communication, I see urging "honesty" as problematic and dishonest. Voting is not an opinion poll; it is an exercise of political power. It's like steering a boat. When you command "right full rudder", it's not an opinion, but a muscular exercise that feeds into the whole dynamic of the boat's motion in accord with the Laws O' Physics (TM), the design of the boat, the propeller's rotational velocity, etc.

Or we could provide true information and help the system pick a better winner.

No, we can't. We don't have access to the true information. We have no means to extract this from the voters. Voters have free will and their own purposes and values. If we are studying what they do, we may indeed get a pretty good clue about their values, but we cannot guarantee to get it accurately. We don't have the power to coerce them into telling the whole truth about it.

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

then I predict that Faction A will within a few elections figure out that it should use the maximum value (5) and minimum value (0) of the permitted range. I think factions don't usually voluntarily give up power. Elections are contentious.

Empirically, about 60% of voters choose to do so.

They choose to give up power? What situations was this empirical measurement made on? Was anything at stake based on the outcomes from the tallies? How many Score elections had the voters already experienced, where something important was at stake?

When asked if they'd prefer to have a voting strategy automatically executed for them,

Again, for what porpoise do you bring this idea into the conversation?

@jack-waugh said in STLR - Score Than Leveled Runoff might not be too complex for voters:

What reasoning leads you to think that Score voters pushing exaggerated support hose toward compromise candidates tends to spoil elections? They are still giving more support to their true favorites. If enough proportion of voters are standing with them, that candidate can win.

Arrow's theorem, which implies that if voters base their on strategic considerations, there will always be spoiler effects.

What is the relevance? The Arrow theorem assumes strict ranking.

(Compare honest score voting, which is completely spoilerproof.)

"Honest voting" is a theoretical concept that can be useful in thought experiments and reasoning and the design of algorithms, etc. However, it does not describe a phenomenon that can happen in real elections in which something important rides on the outcome of the tally.