election mechanics (not about the US)

[cellio]

A few days ago I was musing elsewhere about some online elections. Specifically, Stack Exchange has been running elections to replace all the moderators who have quit, and it's highlighting some weaknesses in their election scheme. Ranked voting is much better than "first past the post" but you still have to put the right checks in place.

If your election system uses ranked voting, think about how voters can reject candidates. The Hugo awards have "no award" as an automatic candidate in each category and you rank all candidates. My local SCA group lets you mark candidates as not acceptable and any who get 35% NA are removed, which gives the voters a veto when needed. Systems in which you pick N candidates lack this safety check.

"Cast N votes" doesn't let you distinguish between "this candidate is ok but not in my top N" and "I oppose this candidate". And even if you allow "not acceptable" marks on candidates (like my SCA group), you still need to allow ranking those candidates so voters can express "the clueless candidate before the evil one". If I recall correctly, my SCA group gets that part wrong; if you vote "not acceptable" you can't also rank the candidate, so you can't express degrees of unacceptability. If your goal is to deter NA votes that's a positive; if your goal is to elect people who are broadly acceptable then it's a negative.

Stack Exchange uses "cast three ranked votes" and now allows uncontested elections, so the only way for a community to reject a candidate is to round up more candidates. Because Stack Exchange royally screwed some things up with its communities, recently there have been newly-elected moderators who'd only been users for a few months. A candidate in one election is largely inactive (and said so).

The new and mostly-inactive users might be fine people, but in the past the bar was higher -- moderators were expected to have been regular, positive contributors for a while. Desperate times call for desperate measures, I guess; SE has lost a ton of mods in the last year for good reasons that still apply, but they don't want to admit there's anything wrong. So it's important to them to have bodies in seats.

Every voting system has flaws. When choosing, you need to decide which flaws are ok, which you actually prefer, and which must be prevented. Ranking all candidates, allowing an NA mark or "no award", and applying an threshold is more expressive than "rank N" but also carries more voter burden. Too complex? Depends on the characteristics of the electorate and the importance of the results, I guess.

Codidact isn't going to mandate a particular election scheme for its communities. Nothing is baked into the software, and on the network we host ourselves, our policy is that our communities can choose their moderators in any way they choose so long as the method produces unambiguous results that can be audited. (That's because any disputes are going to be escalated to us, so we'd better know how to fairly adjudicate them.) But even though our communities can choose how to choose, we should probably plan on offering some sort of facilitated options -- we can run election type X or Y for you, or y'all can do something else. Not every community wants to build its own system, after all; we shouldn't make them. I think we're a ways away from moderator elections yet (our communities are in start-up mode), so there's time to talk with our participants about what makes sense.

Comments

[madfilkentist]

It could be interesting comparing Stack Exchange with Reddit. The Reddit model is totally different. Anyone can start a subreddit. Most languish in obscurity. Success probably depends as much on snagging a title of broad interest as anything else.

The moderators can set their own rules. Sometimes the rules are silly and prevent interesting discussions. For instance, on the Writing subreddit, posts linking to relevant and informative articles are forbidden. Most of the posts are questions by would-be pro writers who haven't learned the basics. Other parts of Reddit are run by petty tyrants who remove anything they don't agree with.

I've been moderately active on Reddit over the years, but I've finally reached the point of giving up. Still, a decentralized model has its points. There must be some really good subreddits, if I knew where they were. Which is better, Stack Exchange's centralized structure or Reddit's anarchy? Both have their points and their weaknesses. Does Codidact draw on the best of both?

[luscious_purple]

I'd heard terrible things about Reddit harassment, so I stayed away for years. Last year, I set up a Reddit account with a gender-neutral-sounding user name (NOT luscious_purple), and I joined a few subreddits that had nothing to do with U.S. presidential politics. I visit the SCA subreddit maybe a couple of times a month, just to see if anybody mentioned something cool that isn't on my usual Facebook groups. I will probably never spend much time on Reddit, though.

[cellio]

Which is better, Stack Exchange's centralized structure or Reddit's anarchy? Both have their points and their weaknesses. Does Codidact draw on the best of both?

I hope we're drawing on the best of both and avoiding the worst of both. On our network we have a community-proposal mechanism, which is much lighter-weight than Stack Exchange's. (Stack Exchange now makes it very difficult to create a community, which fits with both their business model and the fact that they've got 170 of them already.) On Codidact, we'll create a community if -- hand-waving ahead -- there's "enough" interest. "Enough" is a fuzzy mix of number of people, people's specific interests (e.g. if nobody's prepared to answer questions that'd be a problem), and level of enthusiasm. The Judaism community had several enthusiastic people within hours of being proposed; we launched that in a few days. A proposal for role-playing games feels like it ought to have support but people aren't participating much in the discussion so we don't know if we should create it or wait. And, of course, we're new to this and learning as we go.

The reason we have a proposal process and don't just say that anybody can do anything like on Reddit is that we want to support both individual communities and our network of communities. If a proposal would have large overlap with an existing community or another proposal, for example, we want the folks involved to work with each other to figure out how to avoid confusing people. There's nothing wrong with some overlap (and I'd say it's pretty much inevitable anyway), but, for example, while launching our (new) math community we got a proposal for a math-and-physics community, and we suggested that the latter be reworked as just physics, or as natural sciences if the person wanted a broader scope.

We expect communities to have different norms -- about what is expected of questions, about what sourcing or support is expected for answers, about level of chattiness in comments, about how actively people should edit, and so on. Stack Exchange has these variations too, though they might not admit it. From what I've seen, Reddit has even more variation. We have some baseline requirements of all communities, we give moderators broad authority to manage their communities, and we have an escalation path for anybody who objects to an action by moderators or staff. I can't tell if Reddit has any sort of meaningful escalation or review; Stack Exchange's is a joke (they'll set aside their policies when it suits them). We're still at the draft stage, but we intend to give the community a real voice here. We always want the first step in any disagreement to be discussion, working together to find a solution.

(There's probably more I could say but I must run now.)

Edit: somewhat related, on the community-creation part: how granular should communities be?.

[luscious_purple]

One of the things I've learned from a couple of cycles as working as an election judge (a.k.a. "poll worker") is that the voting system should be as simple as possible. If the franchise extends to *everyone*, then it extends to people with developmental disabilities, old people with mild dementia or over-medicated confusion, people who are profoundly dyslexic, people who learned just enough English to pass the U.S. citizenship test and then retreated into their native-language communities after becoming a U.S. citizen, etc. etc. etc.

For example, when I was working at an early-voting site for the 2018 Maryland primary, I saw an elderly woman seated at the "ballot marking device" (the machine that is supposed to make the polls ADA-compliant). She started to scream, "Where are all the Republicans on the ballot?" The judge who was stationed at the device explained that this was the primary, not the general election, and she had been given a Democratic ballot because she was registered as a Democrat. She seemed to understand in the moment, but 30 seconds later she was screaming, "SOMEBODY PLEASE TELL ME WHY THERE ARE NO REPUBLICANS ON THE BALLOT!" It took three election judges 45 minutes to get her to finish voting.

[jducoeur]

Codidact isn't going to mandate a particular election scheme for its communities. Nothing is baked into the software, and on the network we host ourselves, our policy is that our communities can choose their moderators in any way they choose so long as the method produces unambiguous results that can be audited.

In the medium term, assuming you're successful, I'm pretty sure that you're going to have to lay down more guidelines than that -- you're likely to eventually get some petty tyrants who start trying hard to game the system, and there are usually a host of ways to do that.

I'd probably recommend that sometime in the next year or two, y'all do an explicit exercise of, "If I was the moderator, and wanted to retain my seat forever, how could I abuse the system to do that?" Closely related, "If we were a group of loud passionate abusers, how would we do that?" I suspect that will result in some rules around nominations and such -- much of it stuff that looks quite obvious from the outside, but will eventually need to be written down.

(That is, I don't think you need to write a concrete set of rules, but I suspect you will need some regulations around the rules.)