MatrixDevFunnyVideosMusicBooksProjectsAncapsTechEconomicsPrivacyGIFSCringeAnarchyFilmPicsThemesIdeas4MatrixAskMatrixHelpTop Subs
5

With polls, what matters is relative score and outcome quality. I've done two things to avoid some of the pitfalls that befall voting systems that offer a lot of options. Usually voting systems with many options end up selecting for the option that people feel the most passionately about, even if that's reasonable detestment by a large number of people or spite towards that group of people. When given a large list, there are going to be multiple options that could apply to, and so the result is the winner is always something awful. At least more often than not.

The common alternative to fix this is some kind of ranked choice. But that asks a lot of people to order a large list, especially when they may not know anything about some options. For an internet poll, you have to ask for something less demanding. So I've gone with multiple selection, and inclusion of a downvote worth double. So polls here tend to select for a consensus. The winner is something where people like it, and not a significant number of people hate it. If only politics could use such a poll.

The problem is when displaying it. People have a favorite; The most effective way to vote for a favorite is to up vote the few you like and downvote everything else. This isn't wrong actually. A good polling system lets people vote with their optimal strategy. But for displaying it, it means a lot of negative numbers. At the end of the day, it just looks bad.

So we can offset the score upward, preserving relative score, and making a hover over show the raw votes. I was already doing that, making the top item at least equal to its number of upvotes and applying the same offset on everything. But I added another layer that's probably better. Only the bottom 20 percentile can be negative. The 21st percentile and above have to be zero or positive.

You can test it now. Create a poll with !POLL. Downvote half of them. That's more than 20%. The ones you downvoted will become zero and the rest will have a score of 2.

It made the presidential poll look more reasonable. Though that poll shows the downside. Before applying this, everything was negative besides 'no president' and 'Rand Paul', showing that the real consensus when you have a polling metric that measures consensus instead of popularity and cult of personality, that people really want no politicians or libertarian ones (by consensus). Which makes sense. The consensus and common ground is that people want to be left alone. But what's popular is controlling other people. When the poll didn't apply an offset it did better in demonstrating people's disgust with politicians. If 1/3rd or more of society consider you to be harmful, you are likely harmful; And should likely have something nasty done to them if they attempt to exert authority over other people in such a scenario. And the prior one did demonstrate that.

But most polls aren't politics. And polls like movies probably shouldn't be a wall of negative making it look like people fucking hate movies.

IDK, maybe one day I'll find a way to get the best of both worlds.

Update as I'm writing this, I added more information to the hover over, to include the raw score. And looking at that presidential poll, wow, the only thing a significant majority can agree on (2/3rds majority) is that we want no politicians. Rand Paul isn't quite over the line.

One of the main purposes of the polls is to have content that's easily sharable to other sites that does something different than those sites can do. Most sites, even if they have a poll option, only let you pick one item and never let you downvote. They don't find consensus. They find the popular bad option. So yeah, as one of the main ways people see this site, polls kind of need a positive presentation. A wall of negative doesn't do that. And I don't want to blame people for voting optimally and they should be able to do that without feeling bad.

The code:

options.sort((a, b) => b.score - a.score);
favorableOffset = options[0].upvotes-options[0].score
if(options.length>1) favorableOffset = Math.max(favorableOffset,-options[Math.ceil(options.length*4/5)-1].score);

It doesn't apply to past polls unless you click edit, then update. That will re-render your matrixdown with the new code.

Comment preview