Talk:Anti-Harassment Tools/SecurePoll Improvements

It's possible to have an STV system that allows even ranking, and indeed, doing so is necessary in order to handle cases where you've got candidates you'd abstain on (in a normal system) and then candidates at the bottom of the list you'd oppose.

Why has the system made this impossible when the consultation process specifically had numerous statements that an ability to oppose candidates was viewed as needed? Especially when it's not an impossibility? As it is, if I have candidates A-H, I can rank A, B, C normally and I know H is the worst.

A fair (and possible) method would have me abstaining on D-G by giving them an equal ranking and then putting H at the bottom. As this seems written, I would need to just randomly designate rankings to D, E, F, G which is hardly something we want to encourage in an election

@NKohli (WMF): obviously a weekend atm, but just making sure you see this on your return - apologies if double notifying - I've not much experience on mediawiki! ~~~~

@Nosebagbear Firstly, thanks for your comment. Feels good to finally get some community feedback on this important work.

I was not aware that there was a way to implement STV where multiple candidates receive the same ranking. We looked into several STV algorithms and decided to pursue Meek's method since there is a lot of examples of it being used in the wild and I personally never saw one where multiple candidates received the same ranking. My research was limited to reading through relevant wikipedia articles. I'm sorry that my research fell short. I also involved representatives from the Board Election committee in this decision making process but this detail did not get pointed out somehow.

I cannot promise changes to the decisions made at this point. We have a hard deadline to deliver on this project by August 1st and are running behind schedule because of personnel changes on the team. I will see what I can do though.

Can you link me to the consultation process you refer to, if it is public? I have not seen it before.

Thanks again for pointing this out.

Hi NKohli,

Having run actual STV elections in the real world, you certainly wouldn't utilise a tied-vote method there, because manual counting of them would be a massive pain in the posterior, but as indeed Meek's method itself uses, online STV votes have the benefit of using computer calculations. I suspect the secondary source nature of our articles means that our coverage is going to be primarily on the STV variants used in political systems.

The Meek's paper you attach actually notes that you can have an STV method that enables tied voting, though it doesn't explore it since it feels the utility is limited. It's a shame it doesn't - the utility only seems lacking to me if you've got an election where you know all significant candidates well enough to order, but perhaps I'm missing something.

If it is not fixable at this stage, that is vexing, but I guess it is what it is (though whether sufficient Board, ElectCom, and Tech-team time to go through another shift for future elections will be committed, I have my doubts).

In which case the best mitigating solution would be for the built-in summary of voting method at the top of Securepoll, or the lead-in page, to cover the explanation of electoral weighting on this facet - that to give the strongest oppose the system will allow, you need to rank all the other candidates before, even if there are some candidates voters would not normally be able to rank in an informed fashion. ~~~~

In an STV ballot, you should never rank the candidates you oppose with. You should normally only rank the candidate you support. For candidates you don't know, you can either don't rank them at all, or assign them the lowest ranks.

Every candidate who appears on your ballot has a chance of obtaining a share of your vote (say 0.001 for the lowest rank), but I'm sure you wouldn't want to give the candidate you oppose even that small share!