Allow us to rank all the candidates in a moderator election

Question

TL;DR

To reduce the need for tactical voting as far as possible, allow us to rank all candidates in an election and not just three.

Proposal

Instead of just giving us three votes in a moderator election, give us as many as there are candidates (after the primaries). As an equivalent alternative, give us one vote less than there are candidates.

To avoid confusing voters into thinking that they must rank all candidates, leave an appropriate instruction, e.g.:

You can rank as many candidates as you want. If you do not assign the lower ranks, this may lead to your vote not affecting the election as strong as it could. However, the lower the rank the less likely it is to actually affect the result.

Do not label the voting buttons 1st choice, 2nd choice and so on, but rank 1st, rank 2nd and so on.

Finally, as an optional icing on the cake, implement the equal ranking of candidates (which is possible in STV; just split a vote), so I can express opinions like:

A > B > C = D = E = F > G

which means:

• I rank candidate A first,
• I rank candidate B second;
• I cannot decide which of the candidates C, D, E and F is better;
• but they are all better than candidate G.

Rationale

If a voting system awards tactical voting, this is a bad thing as it means that voters who just vote by their actual preference are disadvantaged and the results may be warped due to a complex meta game arising from how voters expect other voters to vote. For example a presumed underdog with enough actual support to be elected may not be elected because people tactically do not vote for them to avoid wasting their vote on the underdog. On the other hand, if there is more than one seat, a candidate who truly is everybody’s favourite may not be elected because everybody counts on enough other people voting for them.

One of the advantages of STV (the voting system used by Stack Exchange) is that it reduces the need for tactical voting (compared to more naïve voting systems). However, as we can only rank three candidates, voting is still more tactical than necessary. For example, I heard people say something along the lines of the following about the current Worldbuilding election (four moderator positions available; a lot of promising candidates):

Candidate X is a certain winner, so I won’t vote for him but rather try to determine the other three positions.

Allowing us to rank all candidates lessens the effectiveness of such a strategy (it cannot be eliminated entirely, thanks to the Gibbard–Satterthwaite theorem) and makes voting less tactical, more fair and the outcome more likely to reflect the voter’s will.

Note that you do not have to have four moderator position for this to make a difference. It already makes a difference as soon as there are more than four candidates.

Also see this question on Politics on how reducing the number candidates to be ranked increases tactical voting in STV.

The only drawback is that the voting process is more intimidating, but a solid explanation should ameliorate this.

A practical example

In the last Stack Overflow election (three moderator positions available), there were 4780 exhausted votes, i.e., votes which could influence the election, if the voters had made an additional choice (which they might not have been able to do due to the limit of three choices). That’s 17 % of all votes.

The last elected candidate (Madara Uchiha) won by a surplus of 144 votes over the first runner-up (Vinod Madyalkar). That’s 3 % of the exhausted votes. So, if only 3 % of the voters whose votes were exhausted had made an additional choice in favour of Vinod Madyalkar, he would have beaten Madara Uchiha in the race for the last moderator position.

(Note that I am aware that this is a little bit simplifying things as all those votes are largely fractional, but the argument is not affected by this.)

Another example is the 2015 Information Security election in which there was only one seat, yet 27 % of all votes were exhausted, a fraction of which would have sufficed to change the result.

"But I do not want to need to understand STV to be able to cast my vote!"

Don’t worry, this proposal is working for your wish, not against it. Tactical voting means that understanding the details of the voting system is necessary to give your vote maximum weight. This proposal tries to reduce tactical voting.

Answer 1

Following our recent update to the voting system, you can now rank all candidates when voting in a moderator election.

Answer 2

The main drawback to any change like this is that it can easily make the voting process more complicated, more intimidating and confusing to the average user that doesn't know anything about STV.

I can't really think of any UI for this that wouldn't be more complicated than the one we have right now. I doubt that it would be a good idea to expose all potential voters to the whole complexity of your proposal. An explanation won't fix the intimidation factor, and I wouldn't really want to change anything that could reduce participation in the elections.

But I think we could avoid most of these issues by leaving the original UI in place for as long as the user hasn't voted for three candidates yet. Once three votes are cast, give the user an obvious prompt to continue casting votes, if they wish to do so.

Answer 3

The simplest way to do this is probably to switch to using OpaVote's own systems not just for counting votes but also for the election. As I understand their pricing structure, an election (where they run the voting UI based on your specifications) costs the same amount as a count (where you pass them the ballot data). I don't know if SE has any discount or legacy billing agreement set up from the old days, but hopefully it would be possible to switch this over without significantly affecting costs. (Elections are, apparently, rather expensive for SE to run already; a modest site with 500 voters might well cost $25 to run the election/count, and SO costs close to $2000 every shot!)

OpaVote's UI offers excellent first-class support for all-option ranking, as one would expect from a professional voting system. There are two similar control schemes to choose from, one in which all choices are included by default and the voter drags and drops them to arrange the proper order, and one in which drag and drop is possible once the voter adds each choice they want to specify. (You can see both in action on their sample poll, contests 2 and 3.) We would probably want the second one, as it allows the current behavior of incomplete ballots in cases where the voter is not sure of their remaining preferences.

For this switchover to work, pricing would need to remain fairly similar, I expect, and rather than switching to the current election-phase codebase when the time comes, the SE software would need to collect and pass along eligible voter emails. This would include one of three (progressively larger) groups:

• Those who visit the page during the election and hit a button to request one.

  Simple. Doesn't bug inactive users. Allows those without email addresses in their profiles to add an email just to vote.

• Those who visit the election page and have the rep to vote when the time comes.

  Even simpler to use, although rather more surprising to anyone who wasn't expecting a mere page visit to automatically sign them up for an email.

• All users on the site with emails.

  Maximum eligibility, simplifying the general process of publicizing the election, and in many cases semi-inactive users would , in the current system, already get a notification that turns into an email.