Policy Review Allow people to rank multiple options equally (AKA tier list-style voting)

[Quanyails]

(Approved by QxC4eva)

Motivation

Many people in CAP have commented how hard it can be to rank options in ranked polls. Some people who can't rank options don't vote until a later poll. This is partly why we have several polls per stage. Allowing for multiple options on the same rank allows people to vote more sincerely.

As an example, consider a user who has a least favorite option but doesn't have a strong preference otherwise. Under the current system, they would need to either:

• Rank every option just to rank that option last.
• Not vote.

This is suboptimal, as neither option indicates the user's true preferences.

Since we moved from Instant Runoff Voting to Ranked Pairs Voting, we rank polls using pairwise matchups. Due to this change, we no longer run into complications if a user wants to place multiple ballots on the same rank.

Concerns + Counterarguments

• Concern: This change allows for tactical voting (e.g. bullet voting).
  • Counterargument: This is possible already through our current voting methods. Any person who wants to vote tactically already can and would.
• Concern: This change makes ballots more complicated.
  • Counterargument: Tier lists are well-known in internet culture nowadays. The idea of tiering options in CAP should be intuitive if not already familiar. Likewise, as this is standard for Condorcet voting method scripts, this change should not require too much additional complexity to ballot parsing.
• Concern: This change makes polls less tractable.
  • Counterargument: Poll results are structurally identical to those of Ranked Pairs, so we should expect no significant differences in how we analyze poll results.

Implementation

1. Update our polling script to handle ballots with equal ranks.
2. Allow for multiple options on the same rank in Ranked Pairs polls. Polls that use Approval Voting will be unaffected.
3. Test this polling method with several low-stakes poll (like dex entries or pre-evos) to iron out any kinks.
4. Evaluate community input and results, then decide whether to keep/revert.

As we want to test this change with dex entries and/or pre-evos, we would like to reach a conclusion by the time CAP 32 dex entries/pre-evos roll around.

----

Proposed revised OP (current, for reference):

This is a ranked pairs poll. The order of your preferences DOES matter. You may rank as many or as few options as you like, but we encourage you to rank as many options as possible. A typical ballot might look like the following:

Most Preferred
Second Most Preferred
Third Most Preferred
Etc.

You can place multiple options on the same rank. If you do, separate each option with a comma (,).

Any comments would go below the votes in non-bold text.

All ballots will be revealed at the end of the poll and votes will be counted via automated scripts. If you do not compose a legal ballot, your post will be subject to moderation.

• Do NOT bold anything other than the options you are voting for.
• Do NOT put any formatting other than bold in your post.
• Separate each option on the same rank with a comma.
• Spellings must match EXACTLY the options listed below.
• Capitalization and spaces are ignored by the counting scripts, but you probably should not depend on it.

Asking for votes for your submission or for the submissions of others is not allowed. Anyone found to have done so risks punishment at the moderation team's discretion. If you find that someone has broken this rule, please contact the CAP moderation team with your evidence and no one else. Mini-moderation of this rule is also considered a serious offense and can be punished.

The voting options are:

Option 1
Option 2
Option 3
Etc.

Those voting options without formatting:

Option 1
Option 2
Option 3
Etc.

This poll will be open for 24 hours.

----

Additional links:

• Discord #prc discussion

 

[QxC4eva]

One thing to keep in mind is that ranked voting is a mathematical paradox (if you want to know why check this link and link). Which means the more you try to "improve" something, the more problems you end up creating at the same time. It's a zero sum game. When you give voters more freedom to vote like that you also simultaneously give more freedom for the paradox to happen, and more freedom for people to manipulate the poll. And you're left with no net advantage in the end. For example:

As an example, consider a user who has a least favorite option but doesn't have a strong preference otherwise. Under the current system, they would need to either:

• Rank every option just to rank that option last.
• Not vote.

This is suboptimal, as neither option indicates the user's true preferences.

If someone wants to vote like that then they're basically manipulating the poll. In ranked pairs the winner is about who has majority support, not who's the least hated one or who's the best compromise between love and hate. Voters who are in it just to bury someone are just as good not voting at all since they're not providing information about who they actually support. If we want to endorse this kind of behavior then we're not suited for using ranked pairs, we'd be more fit for methods that average out preferences without caring about majorities like borda count. For ranked pairs, being able to vote like that is a downside not an up, and it's simply one of the consequences of allowing more flexible ballots. "Good" voters will enjoy more freedom but so will the "bad" voters, and for the good voters the paradox will happen more often and work against them. It all cancels each other out nicely. There's no net advantage.

• Rank every option just to rank that option last.
• Not vote.

These may look like disadvantages at first but if you're following me so far you'd probably guess they come with equally sized advantages too.

• The first one is an inconvenience for sincere voters, that's a disadvantage. It's also an inconvenience for dishonest voters who want to bury something etc, that's an advantage.
• For the second one, keep in mind that ranked pairs is subject to the no show paradox (fails the participation criterion) - which means that some voters can get a better outcome by not voting at all. With the new system, the increase of people voting would be comparable to how much the no show paradox would backfire on them in return.

So again it all cancels out each other. There's no net advantage.

• Concern: This change allows for tactical voting (e.g. bullet voting).
  • Counterargument: This is possible already through our current voting methods. Any person who wants to vote tactically already can and would.

The counterargument is misleading. Any person who wants to vote tactically already can and would - that's true, but they'll have even more flexibility to do so with the new ballot system. As explained earlier.

I'll also note that bullet voting is indeed a problem though we rather people do that than the "reverse bullet vote" example above, where you bury just the thing you hate.

So if it's not clear already, diving into voting mechanics and trying to work against the paradox is ruthless unless one really knows what they're doing. For those that don't but are lucky, the new system has no net advantage or disadvantage at all. If they're unlucky, it would be a worse system overall. The proposal in this PR thread falls into the lucky category. While there are ways to actually improve our mechanics that's beyond the scope of this thread, so let's leave mechanics alone for now and talk about what we can actually talk about - the pros and cons *outside* of voting mechanics. For example, most people in CAP are not aware of just how chaotic ranked voting is, and how this proposal would make it even more so. They probably just see that it'll make voting easier for them and get an (albeit false) sense of freedom or satisfaction out of it. This would be an advantage I can put forward for it.

Disadvantage would obviously be the extra ballot notation and the effort of having to implement/test this change.

p.s. I just want to mention this real quick - if this PR goes through I rather we keep using the older example. This is much clearer

Most Preferred
Second Most Preferred
Third Most Preferred

than this:

Most Preferred
Second Most Preferred, Tied for Second Most Preferred
Third Most Preferred
Fourth Most Preferred, Tied for Fourth Most Preferred, Another Tied for Fourth Most Preferred
Etc.

Especially when most newcomers probably spend no more than half a minute reading the OP before they vote. In that time frame I think it's much easier to infer from the first example how a CAP ballot works.

 

[Quanyails]

I want to highlight the primary motivation behind this PRC suggestion is improving user flexibility, not improving our voting algorithms. Any upsides/downsides to the Condorcet voting results are incidental at best, so I wouldn't focus too much on the theory. If we think this change would seriously damage the integrity of our polls. If so, let me know and we can work things out.

Disadvantage would obviously be the extra ballot notation and the effort of having to implement/test this change.

I've gone through the effort of updating our existing polling scripts to make them easier to develop on and/or test. I'm totally willing to take the bulk of the work, too, so long as I have a reviewer.

p.s. I just want to mention this real quick - if this PR goes through I rather we keep using the older example. This is much clearer
than this:
Especially when most newcomers probably spend no more than half a minute reading the OP before they vote. In that time frame I think it's much easier to infer from the first example how a CAP ballot works.

I updated this section to instead mention that you can put multiple options on the same rank, separated by commas. Let me know how that works for you!

----

Giving this thread another week for input! Feel free to request more time if you have more concerns, though.

 

[QxC4eva]

It's not just user flexibility that improves, but also flexibility for cheaters to cheat and for the paradox to happen. If you're saying the last part is small enough to ignore then the first part too is small enough to ignore. As both cancel each other out.

Thinking about it again, maybe the problem is in our lack of voter education. When people see more options on the slate than they could rank, they probably don't realize all they have to do is keep listing til they get to a rank where they can't, and stop there (even if there's a lower preference they hate more than another). It may seem counterintuitive that they can't 'bury' that specific option but doing so is actually a manipulation tactic, as explained earlier. And introducing tied ranks to ballots would further normalize and reinforce the tactic, in exchange for some perception of flexibility. With better voter knowledge though, voters may be able to see that the "flexibility" is no more than just flexibility to exploit the poll, as burying, sincere or not, plays into ranked pair's weaknesses and should be somewhat happy that our ballot system at least doesn't try to encourage it.

I think I have a stance on this PR now. It's not to go forward with the change but to find ways to show people how to express their vote when they're undecided on some options, which I think could possibly address the issue in the OP.