Talk:Essential Questions

From Electowiki
Revision as of 15:41, 18 August 2005 by Jrfisher (talk | contribs) (Complete Rankings)

Jump to: navigation, search

GENERAL NOTE FOR NON-WIKI-FAMILIAR FOLKS: Please create an identity before editing, and sign talk entries with four tildes (your name and date of comment will auto-generate). Divide categories with two equal signs before and after the heading. James Green-Armytage 02:30, 14 Jun 2005 (PDT)

"Must" vs. "Should"

I'd rather replace "must" with "should" on a lot of the questions... James Green-Armytage 06:34, 11 Jun 2005 (PDT)

I chose to formulate the questions as "sharp" as possible, e.g. using "must" instead of "should", in order to make them as discriminating as possible. I hoped that the distinction between full (++) and partial (+) agreement suffices to distinguish between a "must" and a "should". User:Jobst Heitzig
Or, maybe it could be worded as "should", and a ++ could indicate that the participant feels that it "must"? I've made the change and explained it on EM. Feel free to revert it you prefer. James Green-Armytage 02:30, 14 Jun 2005 (PDT)

Confusing items

I found a number of items confusing:
to make people vote "honestly": Does this mean "permit people to vote honestly," or does it really mean to force honest voting somehow?
to give both majorities and minorities a fair amount of power: What can this mean, other than a Random Ballot component?
Approval information (e.g. cutoffs) should be used: I prefer to get Approval information by using limited ranks, rather than having a cutoff along with a ranking.
Approval information should be interpreted as cardinal rates of, say, 0 or 1: Not sure what the alternative is.
Ranking X and Y equal means X and Y should get the same probability of winning: I get the feeling that this is an effort at describing the WV justification. I'd rather say that "ranking X and Y equal means that neither should get in the way of the other winning."
Freedom of preference expression is more important than anti-strategic properties: What can this mean? What kind of "freedom"? It seems to me that if you can safely express preferences, then this is already an anti-strategic property.
Efficiency is more important than simplicity: Does "efficiency" mean "general goodness"?

Kevin Venzke 20:12, 11 Jun 2005 (PDT)

By "make people vote honestly" I did not mean "force" but rather meant "make it probable that people vote honestly".
A "fair amount of power" need not mean a proportional amount of power as would be introduced by Random Ballot.
Suggesting approval cutoffs was really just an example for approval information, slots could be another, so you could add them as a second example in that statement.
An interpretation of "approved" as "rate 1" would in my view imply that all approved candidates are considered equally good.
The formulation with "ranking X and Y equal" was not an effort at whatever - feel free to add your alternative statement to the list.
As for "freedom of preference expression": It has been stated several times that allowing voters to express, say, cyclic preferences would increase strategic vulnerabilities and should therefore not be allowed.
As for "efficiency", I agree that this term is vage - perhaps we should replace it by "quality of the result" or something along that line.
User:Jobst Heitzig

"Ranking X and Y equal in first-place means neither should prevent the other from winning": How could a candidate do such a thing when they have no power but to vote themselves? Heitzig-j 15:36, 16 Jun 2005 (PDT)

Sorry about the confusion. Consider a method in which the Condorcet winner is always elected. Suppose X and Y are my favorite candidates. It could be that Y would have been the Condorcet winner, if only I had lowered X beneath Y. But since I didn't, someone else entirely is elected. ...I'll try to reword this item. Kevin Venzke 15:54, 16 Jun 2005 (PDT)

Essential Question or not?

I think the question whether "It is preferable to measure defeat strength in pairwise methods by winning votes rather than margins" is important but still not essential in the sense I intended this list to be since it seems to depend mainly on other questions (how to interpret equal ranks, the importance of anti-strategic properties, etc.). Heitzig-j 15:23, 14 Jun 2005 (PDT)

Honestly, it is not clear to me how one's answer to this item can be deduced from other items. Why not have it? There seem to be a number of items whose inclusion seems arbitrary, or which could be broken down into underlying principles. (What's the principle behind approval cutoffs, for instance.) Kevin Venzke 16:32, 15 Jun 2005 (PDT)
Well, I put it there, so obviously I think that it belongs. Actually, I'm quite interested in people's answer to that question. Anyway, I don't see how it's less essential that a lot of the other questions there. It's certainly too general to be asked in the method evaluation poll. James Green-Armytage 01:38, 16 Jun 2005 (PDT)

Complete Rankings

It seems that much of the discussion of strategic vulnerabilites depends on either exploiting or creating ambiguities in rankings. In the extreme, organized bullet voting would reduce Condorcet to first-place plurality with all of its weaknesses. Therefore, I have added a "must complete rankings" (disallow ties and truncation) in segment 3.2, data gathering.

Sure, voters might scratch their heads over a few coin flips, and they may worry about failing "later-no-harm", but each can take consolation from the fact that all of their opponents will be in the same boat. Therefore, the incidental problems should be a wash while the strategies to game the system should be curtailed.

Even better, "attractive compromise" candidates may emerge when polar factions are prevented from concealing their lesser preferences (and when they know that their polar opponents are likewise constrained). In other words, gathering more data increases the odds of making a better decision. Jrfisher 13:29, 17 Aug 2005 (PDT)

In Condorcet methods, the voter who prefers both frontrunners to a bunch of unknown, undesired candidates has three choices:
1. Rank both frontrunners above the unknowns. This is probably unsatisfying for many voters. Also, if the worse frontrunner's supporters don't pick #1 also, this can have the effect of giving the election away to the worse frontrunner.
2. Rank the unknowns above the worse frontrunner, using burying strategy. This can steal the election for your preferred frontrunner, but it can also backfire, causing an unknown candidate to win. But I suspect this way of voting would be satisfying for many voters.
3. Truncate, and don't rank either the worse frontrunner or the unknown candidates. This makes choice #2 more likely to backfire for other voters, so that they won't want to try it. Truncation is also easier, and doesn't take a lot of thought to do.

I am very worried that if you disallow choice #3 in a Condorcet method, voters will always have to play chicken between choices #1 and #2. Kevin Venzke 14:57, 17 Aug 2005 (PDT)

I see your point, but I think it far less important than the polarization (center squeeze effect) that comes from truncation, especially when taken to the extreme of bullet voting. We could enact a Condorcet system only to see the two-party racket persist for yet another century or two. User:Jrfisher

I don't think it is less important to avoid electing candidates no one wants. It will be a major obstacle to get Condorcet methods adopted if it appears there would frequently be incentive to bury, and that it could frequently backfire. Kevin Venzke 07:12, 18 Aug 2005 (PDT)

Besides, voters need not always play chicken between #1 and #2. They could learn enough about the unknowns to render them known so that wise (sincere) votes could be cast. Indeed, disallowing truncation could be the very spur we need for that beneficial outcome. It would also motivate all frontrunners to compare themselves and each other to all outside challengers instead of acting as if the others weren't even on the ballot. User:Jrfisher

No, informed voters make no difference. If I didn't call them "unknowns" but "undesired candidates that no one wants," the only difference is that choice #1 is already sincere. Actually, I was assuming that #1 is sincere when I wrote the above. Real compromise candidates wouldn't necessarily make a difference: As soon as they're perceived to have a chance of being elected, they will be a target of burial by voting method #2. Kevin Venzke 07:12, 18 Aug 2005 (PDT)

Thus, the problem raised is a "feature", not a bug. Voters have an outlet and it is beneficial if they take it. After that, if we are still worried about "nonsense" candidates creating tedium, then we have a ballot access issue that can be solved separately. User:Jrfisher

I'm not talking about "nonsense" candidates, simply candidates who are not perceived to be viable. I don't know how you can modify ballot access so that all candidates must be perceived to be viable! Kevin Venzke 07:12, 18 Aug 2005 (PDT)

However, if we are still worried that the electorate is too lazy to figure out how to sort a ballot, then Condorcet is already doomed. If we design a system around voter ignorance, we end up with something very much like the wretched system we already have. User:Jrfisher

One problem is that when voters can truncate in a Condorcet method, they usually should at some point, probably above the worst frontrunner. I think the desires are basically incompatible, to want a Condorcet method, and to want voters to rank all the candidates. Kevin Venzke 07:12, 18 Aug 2005 (PDT)

Also: Though serious students of voting methods may have decided that winning-votes (WV) are more usable tie-breakers than margins, the public may find the familiarity of margins too attractive to let go. According to Armytage, disallowing ties (hence truncation) eliminates the distinction between WV and margin. That would remove one obstacle to adopting any of the strength-of-defeat methods.

He's correct. But James Green-Armytage also believes margins is completely unusable for public elections. Doesn't it say something, that barring truncation makes the usable method equivalent to the unusable method?User:KVenzke
Yes, it says that the problem with margins is created by truncation and tied ranking. Jrfisher
I notice that James doesn't even comment on whether barring truncation would actually be a good idea.User:KVenzke
If the public is drawn hopelessly to margins, then I would say Condorcet methods using pairwise defeat strength are doomed. But I don't think they will prefer margins, since it adds an extra calculation to find every defeat's strength. If a method definition says to use the literal number of voters voting a certain way, I don't think it is so intuitive that random people in the public will say, "Wait, why can't we subtract another figure from that figure?" Kevin Venzke 07:12, 18 Aug 2005 (PDT)