I think that this thread is a great example of why we should adopt a simpler voting algorithm (like approval voting or score/range voting). It shouldn’t be this hard for the voters to reason about the outcome of the vote.
19 Likes
Do you mean general GitHub flakiness or something else? Because there was access to the collection of the candidates’ statements by candidate in the candidate files, and you can look the issue discussion to look at it by question, I am just not sure what «objectively definable» view you mean.
Reading this volume of statements is indeed inherently inefficient, but unfortunately any improvement on e.g. similar but also meaningfully different questions involves judgement calls that cannot be really entrusted to anyone globally/officially.
Any proportional system has to have somewhat complicated behaviours.
Any reasonable non-proportional system is Condorcet-ish single leader election with some constraints on expressible ballots plus extra overhead, we can as well skip the entire committee part then.
7 Likes
Not the original objector, but I found the formatting of the candidate files to be particularly bad. Having to click on each of the questions to expand them was tedious, and once they were expanded, there were insufficient formatting clues separating question text from answer text, particularly given how long some of the question texts were, how much they used their own headers, and how (some) candidates used (a variety of) headers themselves. It was not conducive to easy skimming.
To elaborate, for those who agree with @Gabriella439 on this: If there is a popular candidate A, in an approval-voting election for 5 seats, A’s best strategy is to encourage just 4 other people to run who will say that they agree with A on everything, and to put out a ‘voting guide’ encouraging everyone who supports A to also support the clones. The electorate has no reason not to do this if they want to maximize their favorite candidate’s impact. Regardless of what the other candidates do, if A gets the most approval votes, so will all of A’s clones, and therefore only A and their clones get elected. So why bother with a committee? Just elect A and let A do what they want. Score/range voting is similar.
Approval voting is great for maximizing satisfied electors but not suitable for representing minority viewpoints, and if you don’t care about representing minority viewpoints, just let the most popular leader appoint a cabinet and have fewer arguments.
15 Likes
I don’t think there’s much confusion in this thread: there are some people poking at the data, but I read that as curiosity more than anything else. One person asked why OpaVote displays incorrect results, but there would be similar confusion if seven people were displayed as winning a five-winner approval-voting election, and the obvious solution is “configure the number of winners correctly next year”.
8 Likes
Thanks @ners, @7c6f434c and @RossComputerGuy for running this election process.
It seems that many candidate’s focus is on improving the SC process itself, which should lead to more transparent and faster decision making from this body. I’m looking forward to see that happening.
20 Likes
Yes, that is what I had in mind.
Ah, I wasn’t actually aware of how the answers were collected. I agree that helps a lot (link).
Some rough ideas:
I’m thinking some spreadsheet/table would be nice, where I can focus on / filter / expand rows and columns.
So that I could read e.g. along a row to see what candidates think about questions I care about, and then e.g. for candidates with whom I agree, read along columns to see whath their other opinions are.
Or maybe the table should just have an icon in each cell where a candidate gave an answer, and then I could click it to read that answer in a modal.
Even better would be if I could then assign a numerical score to that cell, so that after reading all answers and judging them, I could look at the table summarily to order my voting choices.
Not sure if anybody already made such a tool.
1 Like
Oh, interesting. (And probably good, and also ideally needs a few iterations over a month-ish).
Thanks for the detailed description of your UI ideas!
These are the things that would definitely benefit from SC first deciding the platform-dependence position for the next year, then appointing a standing EC, which would have the authority to approve-until-future-notice a tool like that for future use.
Yes, by an actual election the EC could rotate completely and also overrule some of the previous stances. But I think not-intended-as-political teams (SC and mod team are explicitly dealing with politics) default to keeping the direction (even across rotation) unless there are reasons not to. And the tool needs static data as the input anyway, and having some way of Q&A is natural, so the tool should be safe.
1 Like
Id like to chip in here, that I’m in favor of the current system. As said it maximazes voter satisfaction and let’s me vote (even though I couldn’t actually vote this time) honestly. Having to vote for someone I don’t 100% like just because I think they’re the most popular candidate is not great.
If you’re at a loss how the system works, see https://www.youtube.com/watch?v=l8XOZJkozfI which does a good job at explaining STV. (I would like to ask the EC to verify that we are indeed using the STV system depicted in that video, until that happens take my words with a grain of salt).
6 Likes
This system, while a bit complicated, has seemed to give a result that helps give a broader spectrum of perspectives and representation in the elected SC.
Now, of course, the hard part in keeping that diversity is that the elected officials will have to show up every time to do the dreary and dull work of participating and voting. But, I believe in boring Nix, and I believe that the new folks will be more than capable in this regard, hopefully even surpassing the previous SC in attendance and participation!
And for folks concerned about this, @Infinisil has been keeping excellent SC meeting notes (check here, I believe), and you can see how the people you voted for are contributing–or being absent.
7 Likes
Congrats to the new SC members! Looking forward to working with you as part of the NixOS Foundation board 
The one major improvement I see for future elections is to make candidate evaluation less time consuming. I had the luxury of being able to go through more or less everything the candidates said, but even then I had to skim many parts, and this took me about 4 hours! I know some voters with busy lives with spouse and kids, who told me that they struggle to dedicate any time at all towards this, so I think we should improve this.
Concretely, some ideas are:
- Set strict limits on one or more of:
- Candidate statement length
- Who can ask questions
- Number of questions per voter
- Question and answer length
- Increase candidate requirements for an easier selection, e.g.
- Increasing the number of required endorsers or contributions
- Having a pre-selection round
- Introduce more efficient ways of getting a sense of candidates, e.g.
- A well-reviewed standard “agree/disagree”-style questionnaire for all candidates, then visualising the answers in a compact form
- Interviews for all candidates, with the same questions for everybody and an unopinionated moderator, with strict answer time limits
I think we should strive to make it possible for somebody with just 30 minutes of spare time to get a good sense of all candidates. This will also improve voter turnout, which ideally would be 100%
33 Likes
I would absolutely appreciate a reduction of text to read. The amount of information available while the voting period fell into family holidays made it hard for me to keep up and I didn’t cast my vote at the end, because the last couple of hours left in the voting didn’t feel enough to build an educated opinion about each of the candidates.
5 Likes
As a perfect example, I just stumbled upon this blog post from @patricksjackson, who only had about ~1 hour during vacation:
I’m of course not a fan of LLM’s having such influence, even if what it output was verified as Patrick states, but we shouldn’t bash people to reach for such methods considering the necessary time investment, and instead aim to decrease the amount.
4 Likes
There were four weeks of time between closing the applications and closing the voting. I agree that 1h isn’t much to make a good choice unless one had significant knowledge about at least part of the candidates from before.
Ultimately it’s up to each individual what they do with their votes. It’s practically impossible to enforce a ban on this even if we wanted to.
6 Likes
I think that the discussion over the amount of information per candidate to consume is important, but it misses an important other aspect that heavily contributes to the amount of effort required for voting: the fact that we have to rank all candidates against each other is inherently quadratic in complexity. Combine that with roughly half of the candidates being mostly unknown and a whopping two dozen candidates to pit against each other, and no wonder that people resort to using opinionated voting guides or LLMs to reduce the workload
15 Likes
You don’t necessarily need quadratic mental complexity. You can e.g. put them into three buckets (good, bad, I don’t know) and then sort each bucket randomly. It’s everyone’s tradeoff how much work they spend on details (or if they get help from LLM).
I even did something similar, though the randomly sorted buckets were small, 1–4 members in each.
6 Likes
I disagree with introducing stricter requirements.
I do agree however that a very basic comparison tool would be good – something like voting assistant software for national elections; where you punch in your opinion on important issues, and it shows you with which candidate’s views your opinions align the most.
One for example (first English example I found while searching, don’t know anything about them) is euandi.eu.
This is completely developable within the question system we have right now – seeking an official “how much do you align with this statement, 1-5” response from candidates would help when creating such a tool, but it is not a hard requirement.
Of course, such tools should be used as an aid to, and not instead of, the official candidate statements.
I believe that introducing stricter regulations for candidacy necessarily reduces democracy, and how representative the SC is of the opinion of the community.
4 Likes
Completely agree. The problem here seems to be people having to make optimal stopping decisions with limited time. Cutting off information is a false solution: it makes it easier to make a decision at the cost of the actual usefulness of that decision.
And I also totally agree with your solution.
In Denmark, for Municipal and State elections, the state television (think Denmark BBC) produces a quiz that lets all candidates answer on a scale from 1-5, disagree to agree about certain issues.
The trouble with such tools is that those that create the questions have an outsized control on the bias of the questionnaire.
I think a pretty cool solution would be something like this: In advance of the election, potential questions are collected from the community as a freetext field. Then, the community votes for the questions they want to see the most. Then we take the 20 most requested questions, and make part of the application for candidacy to give an answer to the scale.
This does introduce the problem of introducing false dichotomies, or turning something qualitative into something quantifiable. The solution the Danish state media went for here is to have a mandatory 1-5 rating, but an optional free text field for the candidate to explain themselves.
I think something like this would work exceptionally well for helping people make quick decisions, without restricting the amount of information in circulation.
I’d love to hear how @Infinisil @NobbZ @piegames as well as others feel about a solution like that?
9 Likes
i agree that centralizing questions would centralize power in the ability to frame the debate.
but if people care more about some questions than about others, one could just focus their candidate research on the questions they care about most. why then cut off someone else’s access to questions they might care more about?
2 Likes
Every candidate is free to have a concise pitch as the first paragraph of their text, done in whatever way they prefer. It’s in their interest. Perhaps for some people it will suffice to read just that (20-ish times) to decide, or they may read more.
2 Likes
Hard to say for how much shaping EC has legitimacy. Although of course a constitutional grant could solve this. (Standing EC could also have time to judge the reactions to various options).
One could apply to questions same scrutiny as to candidates for example, author+endorsers should have 4 non-mutually-interest-conflicted people among them.
Definitely amendment territory (and loses representation of combined positions, I guess)
Sorting is N log N.
This is not an empty quip — I did try to skim full statements + answers, and extract key points, and then sort everyone carefully, but I did basically learned-insertion sort, look where in the ranking the positions «feel like», then compare in the small neighbourhood.
To be honest, in a proportional system it is pretty reasonable to put at the top people who you actually trust, and in particular somewhat know (and in the bottom half it is fine to shuffle groups that you don’t know how to compare)
A moderator that is trusted by a supermajority as unbiased for real-time interaction sounds like an extremely high bar!
I feel like any influence over «content» of the discussion needs a constitutional grant, and not sure how to do it properly/safely.
Yeah, that’s why I did not want EC filtering questions without being explicitly externally tasked with that.
This omits an important part of the issue. We already had similar questions with various degrees of difference. So we need to decide «which questions do not belong together», «which of the same-niche questions is better», and «which question groups are more important».
Although maybe entrusting EC with bucketing only could be low-controversy-enough.
Very much valuable, yes!
I think a reasonable compromise is something like:
- bucket (judgement) and sort (poll) questions
- the candidates are «expected» even if a failure is not technically disqualifying to give strong-agree-to-strong-disagree score or «objection to premise» to winning questions in top-N buckets (with optional freeform comments for the entire bucket, in the table it is «click-for-details» or lack thereof)
- but one can answer other questions, and voters are free to scroll the table right.
2 Likes