Election voting systems (STAR, proportional, etc), Bylaws

First, this is a huge topic, so to reduce the likelihood of this growing out of proportional (hah pun!), note that active theory and advocacy discussions are at https://forum.electionscience.org/ and https://www.reddit.com/r/EndFPTP/ and elsewhere. If you want to have fun diving into these things, check out this animated game thing: https://ncase.me/ballot/ (fully FLO!) and this derivative (hooray for FLO!) https://paretoman.github.io/ballot/newer.html

The questions for us are what to do here for our co-op. How specific (what features to lock-in at the Bylaws level), etc. I’m here summarizing the thoughts about this in order to get feedback and questions. Please make generic theory questions into new topics in #general-discussion (or take to the forums listed above). This topic itself should focus on discussing the decisions for our co-op specifically.

The all-around best voting system for single-seat elections is STAR (Score Then Automatic Runoff), which is supported already by fully-FLO polling software at https://star.vote (code at https://github.com/msmunter/star.vote/). Trying it yourself is there is the easiest way to see how it works. In short: scores of 0-5, sum the scores, of the top-two candidates check which had more ballots showing preference. It can be thought of as a primary and general election all-in-one.

Proportional Representation (PR) is great for multi-seat elections. It adjusts the weighting of ballots to aim for as many voters as possible to get representatives they support. Several systems are possible, including using the same 0-5 scoring ballots as STAR. However, all PR systems are relatively complex compared to single-seat systems or non-PR multi-seat elections. PR systems are less supported by online polling tools currently.

In meeting with @board we approved of a great suggestion from @lishevita to allow the Board in the future to designate named seats (generally for the purpose of promoting diversity or particular underrepresented concerns). They could be named for an honored person, for example. They would not have any requirements, so such seats would be open to any candidates. But they would have a social pressure. If a named seat signified somehow (for example) non-English-speaking and lower-income people, then it would seem wrong (but not be disallowed) to elect a wealthy American monolingual person to that seat. Our thought for the Bylaws is to explicitly allow this sort of tool without mandating it. That way, if it seems worth using, members could not object that this tool is unfair or against the spirit of the Bylaws.

Bylaws wording drafting

One option would be to pick specific systems. So, for example, we’d say in the bylaws that elections shall use STAR. However, to keep in more flexible, we could mention the core principles that matter. So:

“Preferential voting: allow (but do not require) voters to express preferences across all candidates”

In other words, don’t limit voters to pick only 1 (or only 3 for 3 seats) out of a pool of 10 candidates. At the simplest and lowest-resolution, this includes Approval Voting, which is simply the ability to vote for any number of candidates. Approval is better than limited-choice voting but much less expressive and robust than STAR. (Side note: OSI uses approval voting via https://vote.heliosvoting.org/ — disclaimer: they requested my advice when they initially made that decision)

“considers all marked preferences in determining the outcome”

In the meeting, we didn’t reach consensus on this wording, but decided to consider it after further discussion. This criterion fits several systems, STAR, plain score, approval, 3-2-1 and others.

This would disallow IRV (Instant Runoff) (and its proportional version STV) which is the only version of Ranked Choice voting in active use anywhere. Contrary to most lay intuitions, IRV discards a lot of the preferences and that leads to problems. With IRV, a super popular candidate can get eliminated due to lack of 1st choice votes (which are the only ones counted initially), and then all the 2nd choice votes for that candidate get ignored — while some other voters get their choices counted. To understand this clearly, see https://www.electionscience.org/library/the-spoiler-effect/

One way this discarding of markings shows up in IRV is “favorite betrayal”. If there’s a risk of eliminating your 2nd choice if your 1st choice gets a bit too strong (but not strong enough to win), voters have an incentive (which gets clearer the more experience they have with IRV) to “betray” their 1st choice and put the “safe” 2nd choice as their 1st choice on the ballot.

“use proportional representation for any election with multiple open seats with the same designation”

As we like PR, we could lock it in. However, it’s not compatible with designated seats.

Also, if we use designated/named seats ever, we wouldn’t want it to become tokenized. So, we wouldn’t want all the candidates for a designated seat to be excluded from the rest of the pool of seats. So, we need some process that combines these things if we keep that idea. It seems, we’d have to have full a single-seat vote for the designated seat, and all the candidates for that seat would be included in the at-large election; and the winner of the designated seat would be eliminated from the at-large pool before calculating the winners there (in PR, that elimination could be used for reweighting the same as usual). This is getting complex but is particularly flexible and nice combination of ideas…

Yes/no ballots

We’ve had the idea for a long time to use a range (like Likert style, strongly approve, approve, slightly approve, etc.) instead of simple yes/no. This extra expressiveness could allow for passing measures that have a lot of weak skepticism but strong support or rejecting measures that have only weak but widespread approval and strong minority opposition.

A connected idea: use some combination of median and mean to determine something like “supermajority”. So, for highly-important things (perhaps like changing the Bylaws), we want not only 2/3 of voters approving, we’d also want the average level of approval to be high…

Thoughts, questions?

Clearly, many of these ideas might be best put into a list of ideals that we plan to support when we have the bandwidth. Perfect is the enemy of the good.

My feeling is that awareness of and thought about these types of democratic processes matters a lot. Understanding the impact of the systems is a start. Whether we settle all this now, I hope for us and for the wider world that more people be interested in the questions of how to best make collective decisions — the core stuff of cooperation and coordination and the Snowdrift dilemma etc. Many if not most things we care about in the world are most affected by how and whether we can work together effectively.

Hoping to get others engaged on these things still. Notable update in my thinking:

Lottery options?

Snowdrift co-founder @david has been long mentioning the idea of Citizen Assemblies (from before he knew the term). But I was recently prompted to consider this sort of thing for co-ops by a recent Revisionist History podcast on elections by lottery.

Skill at campaigning does not correlate with skill at governing.

Perhaps we should consider or otherwise at least allow the option of having representatives (and/or perhaps committee-members) elected by:

  1. narrowing to a pool of willing and qualified (what qualifications?) candidates
  2. selecting from that pool by random lottery

We could designate a portion of seats on Board or committees as at-large seats selected by lottery this way.


Another (compatible) option to avoid biases in elections is to only reveal select things about candidates so that voters are choosing about values, policies, qualifications etc. and not someone’s looks or charisma or reputation. I’m not sure about this sort of thing, but I see some merit to considering it.


On terms for honorary seat designations

One updated thought on this point: perhaps we want it more clear how this might work over time. So, I wonder what’s best in terms of whether designated seats come and go or get re-designated to keep rotating the focus to establish diverse diversity (if you will). One-term designations vs indefinite? Specifying designations as one-term with the option to renew would make changes less controversial than having indefinite designations that get repealed/changed.