LJ Advisory Board election ends TODAY. Please VOTE!

[elements]

[It's been a long time since I updated - I'll post later about what's been up with me but yes, I'm still alive.]

The election for user representatives to the Livejournal Advisory Board ends TODAY, at 9pm Pacific/midnight EST. Please, please, take the time to vote.It's important that LJ sees strong user participation in this election, and we have a

Comments

Why this "instant runoff" is broken

[cos]

As you know, the idea behind instant runoff is that in each round, as you eliminate candidates with the fewest votes, you slowly converge on a candidate who a majority of voters feel okay with. However, when you have 23 candidates and are only allowed to vote for 3, most people can't vote for all the candidates they're okay with, and worse, most candidates won't get the votes of all voters who are okay with them. That makes it much less likely that we can converge to a majority: ballots will be "exhausted" too quickly, not because the voters didn't have other candidates they might have approved of, but simply because there wasn't space on the ballots to list those candidates.

Adding up the number of first round votes right now, I get 25,346 voters. Assuming that everyone voted for 3 candidates (which I think is close to the truth), that means a candidate needs 12,673 votes to win. The current leader is legomymalfoy and if you add up all of her first, second, and third round votes you get 13,217, just barely enough. That means that depending

Re: Why this "instant runoff" is broken

[jiggery_pokery]

See, I've been saying exactly that. (For instance, bookshop quoted me saying that here.) Thing is, I'm now convinced I was wrong to say that - so you're wrong to say that, too. :-)

The reason why we were both wrong is that we haven't taken into account that a candidate doesn't need to get 50%+1 of all votes cast, they just need to get 50%+1 of all votes remaining - and, once a vote has all its candidates eliminated, it drops out of contention

Re: Why this "instant runoff" is broken

[cos]

No, I explicitly said "Assuming that everyone voted for 3 candidates (which I think is close to the truth)". If a significant number of people vote for less than 3 candidates, that may very well be because they realize that this system is broken, and only the first round matters, so that might make it seem to work, but would only be further evidence that it is broken.

If, on the other hand, people mostly do vote for 3 but the system only "works" because ballots are exhausted early in the process, such that the number needed to win shrinks rapidly... that is another terribly dysfunctional way this election fails: those ballots exhaust early not because the voters didn't favor more candidates, but because there was no room on their ballots to list them. We still wouldn't get a legitimate IRV result.

Re: Why this "instant runoff" is broken

[cos]

I just looked at the results and this is exactly what happened: legomymalfoy stayed within a few hundred of her initial 9443 through most of the rounds, until the last few rounds started exhausting ballots at a rapid clip. Round 1 had 25875 votes, with lego leading at 9443; Round 15 had 24,393 with lego leading at 9865. Then in the last 6 rounds, the number of votes dropped all the way to 19,445, and lego had 12,630 of those, a little less than 50% of the initial total.

So, first: this only "worked" because so many people abandoned their favorites and voted for a perceived winner, giving her a first round lead that mostly didn't change,

And, second: it depended on rapidly exhausting ballots in illegitimate rounds to bring a candidate to 50%, which turned out to be equivalent to stopping the rounds at the point where they'd still have been legit and falling back on the initial winner.

So many ways this was broken!

Re: Why this "instant runoff" is broken

[jiggery_pokery]

I absolutely agree with all you have said in your two replies to me; it's entirely possible that there could have been a candidate out there who people could have wanted to list ahead of (at least) their actual choice for third preference but felt that it would have been a waste and so did not do so. My comment was merely to dispute the argument that the IRV part of the process wouldn't throw up some sort of winner, even if (as has quite possibly happened) one where the votes cast by the voters did not necessarily reveal their true preferences but ones forced upon them by the system