reconsider STV

[TzviyaSiegman]

STV was introduced to help diversify AB and TAG in 2017. I think the experiment has run its course. While both AB and TAG have (at times) had more gender-balanced, regionally-balanced, and industry-balanced representation, it is not clear if this is because of STV. It is hard to distinguish between correlation and cause. We SHOULD look at other ways to have more diverse elected officials (specifically, we should have a more diverse organization).

Many people strongly object to STV, and many people simply find it an unpleasant process. I think that newcomers find it difficult to understand. Perhaps the time has come to review https://www.w3.org/2020/Process-20200915/#AB-TAG-elections.

See also https://github.com/w3c/w3process/issues/60 https://github.com/w3c/w3process/issues/115

[dwsinger]

I think if we are to reconsider we will need roughly what Chaals did when we adopted it, plus a bit:

1. propose a specific alternative, and explain why it's better and still meets our goals
2. work out a way to trial it, so we learn
3. convince people (AB, team, AC) to support it.

On the first, as you say, the purpose was to increase diversity, so that will need addressing. For example, we could show how the proposal also supports diversity; that while diversity is important, we have the wrong 'scale' of election for STV to achieve it (too small an electorate, too many seats), and so on. I doubt we'll want to argue that diversity is not, per se, a desirable goal, but maybe there is a better way to think about it. (For example, approval voting tends to select candidates that the most feel will 'adequately represent them', and maybe that, or some other goal like that, is what we need.)

[dwsinger]

To be clear, I am not opposed to changing the voting system. I just want to know what would be the proposal, specifically, and why it's better. I'd also suggest that before it comes to the CG, getting consensus on the direction, at the AB, would be prudent. Generally we expect to implement, but not make, "big policy" decisions, in the CG, and voting is pretty central.

[mnot]

My .02 - any proposed replacement needs to be back with research as to why it's superior, just as was the case for STV. I strongly suspect that a driving factor for many of the objections is that Americans aren't comfortable with it -- a poor reason to change.

[frivoal]

I'm not American, and my dislike for STV is not due to discomfort, but rather primarily due to the fact that the specific type of diversity that STV is good at supporting is one that is just as easily described as factionalism, and that this is a poor fit to elect people to bodies (AB and TAG) that are expected to work by consensus, unless the goal is to make those bodies ineffective.

STV elects uncompromising candidantes who are loved by a (potentially) minority group, even if they are strongly rejected by everyone else (potentially even by a super-majority), over candidates who are interested in and capable of seeking compromise and have broad but not necessarily passionate support. When used to elect something like a parliament, where matters are decided by voting, this isn't a problem, as uncompromising minorities can be heard, but not block everything. When electing a consensus based body where any disagreement isn't far off equivalent to veto powers, promoting uncompromising minority view points is not constructive.

Moreover, the various constituencies that STV empowers are, in the case of W3C, much more about industry verticals than about the axes that are more typically considered when thinking about diversity. Some measure of diversity on this topic is good too, but there wouldn't be 9 or 11 browser vendor representatives elected to the AB anyway, because there doesn't exist that many. STV is probably good at supporting geographic diversity, particularly for countries that tend to feel isolated due to linguistic or cultural barriers, but looking at the AB's reasonably good track record on this, it predates the introduction of STV, and the TAG's reasonably poor track record on this hasn't been visibly improved by STV.

For the TAG even more than the AB, where there tends to be an explicit desire by many voters, who are not necessarily themselves diverse in the relevant metrics, to see a diverse group of (cooperative) people elected so that problems can be looked at from many angles, STV does not enable voters to express that preference, while something like approval voting would.

I suspect this wasn't readily apparent in the research we looked at because most research seems concerned with elections of public officials (parliaments and the like), where the factors are quite different. This isn't to say that research is irrelevant. But I would not put too much weight in research papers say that one or another system is good or bad along various metrics, without being particularly careful about what assumptions they had about the electorate or about the body to which people are being elected.

[palemieux]

For the record, can someone provide a reference to studies, theoretical and/or practical, on the use of STV where:

• the size of the electorate is on the order of 100 or smaller
• the number of candidates is on the order of 10
• the number of seats is on the same order as the number of candidates

[chaals]

@palemieux I think you'll find as much info as you need by looking at WikiPedia's articles on voting systems.

@frivoal wrote

STV elects uncompromising candidates who are loved by a (potentially) minority group, even if they are strongly rejected by everyone else (potentially even by a super-majority), over candidates who are interested in and capable of seeking compromise and have broad but not necessarily passionate support.

This is not correct. STV is not influenced by whether a candidate aims to compromise and work collegially or not. It is influenced by what the voters actually want - and modern STV counting systems (such as Meek, the one we use) are designed to ensure the views of the majority have some impact even on the choice among the candidates whose support only comes from a large enough minority to get elected.

STV can allow a very focused minority to elect a given candidate - as you note for its support of geographic diversity. But the fact that the majority's view is taken into consideration means that unless there is a very focused minority aiming to support a single candidate, the majority's opinion of several candidates supported by the same minority has an influence on which of them is chosen.

[chaals]

It is true that splitting an election into two, as we currently do by staggering terms, reduces the ability of any voting system to support diversity. This is the case under any voting system, even approval (which generally works against diversity). The purported benefit of staggered terms is to ensure that continuity provides some ongoing measure of institutional memory, rather than finding that everyone is new. That is a separate question, and I will raise a separate issue to discuss it as a proposal.

[palemieux]

I think you'll find as much info as you need by looking at WikiPedia's articles on voting systems.

@chaals All the examples on the wikipedia page are for large electorates, i.e. county, country... Surely, some references for the use of STV in small groups must have been presented the last time around?

[chaals]

@palemieux sorry, I don't have specific examples to hand. I assumed that people can work small examples by themselves - the system works basically the same way whether there are 10 voters or 10000.

We also ran experimental STV votes alongside several elections before adopting STV. I'll leave it to Jeff and the AB to work out what information to give about the different outcomes that would have resulted if the STV vote had been binding at the time. (My preference is to provide more information than at the time, given that there are no sitting members whose legitimacy can be challenged, since nobody is now currently elected by only one of two parallel votes).

[michaelchampion]

Strongly agree with @frivoal :

the specific type of diversity that STV is good at supporting is one that is just as easily described as factionalism, and that this is a poor fit to elect people to bodies (AB and TAG) that are expected to work by consensus,

I would strongly support efforts to use STV more widely in the US, which is already highly factionalized and where majority voting (not consensus) is used by the elected bodies to make decisions: STV helps ensures fair representation of the various factions, and their representatives can align with other factions to create or block majorities on different issues.

But it seems clear that we need some sort of consensus-seeking process to select people to serve on bodies that are supposed to operate by consensus. Alternatives to Meek STV as a way of selecting the bodies that will work by consensus include:

Nomination Committee -- https://www.ietf.org/about/groups/nomcom/, https://westsidetoastmasters.com/resources/roberts_rules/chap12.html Approval Voting - https://en.wikipedia.org/wiki/Approval_voting Range Voting - https://en.wikipedia.org/wiki/Score_voting

Or some combination, such as a NomCom with an AC Approval ballot. I haven't done a lot of research on this, but I'm particularly intrigued by Range/Score voting as sort of a middle ground between the the traditional voting method W3C voting method, STV, and approval voting.

An alternative would be to accept factionalism as a fact of life, continue to use Meek STV as a way to select decision making bodies, but adopt majority voting as the decision making principle in the TAG, AB, AC, etc. That seems unlikely to produce credible industry standards that the losing side would implement.

[cwilso]

The experimental votes we ran with STV were, IIRC, not particularly applicable to our situation today, as we tend to have many more candidates compared to seats. I agree that splitting our terms across two years means STV doesn't apply its goals as well as it can, but disagree on the effect - in fact, factions can frequently get TWO candidates elected (one per cycle), rather than one.

@chaals I have to disagree that the effects of STV were well-understood (or even presented, in pragmatic terms) prior to adopting it as a system. Saying "It is influenced by what the voters actually want" is a radical oversimplification - ANY voting system can be said to do that. In the end, I find myself against STV because it promotes factionalism, requires antagonistic campaigning, and devalues thoughtful, centrist candidates in favor of factional candidates. I continue to be aghast that we did not implement "no other candidate", as well, as I find that a critical backstop to STV.

Regardless, I wanted to explicitly express my support for approval voting, and my disapproval of STV.

On Tue, Dec 22, 2020 at 9:45 AM Michael Champion notifications@github.com wrote:

Strongly agree with @frivoal https://github.com/frivoal :

the specific type of diversity that STV is good at supporting is one that is just as easily described as factionalism, and that this is a poor fit to elect people to bodies (AB and TAG) that are expected to work by consensus,

I would strongly support efforts to use STV more widely in the US, which is already highly factionalized and where majority voting (not consensus) is used by the elected bodies to make decisions: STV helps ensures fair representation of the various factions, and their representatives can align with other factions to create or block majorities on different issues.

But it seems clear that we need some sort of consensus-seeking process to select people to serve on bodies that are supposed to operate by consensus. Alternatives to Meek STV as a way of selecting the bodies that will work by consensus include:

Nomination Committee -- https://www.ietf.org/about/groups/nomcom/, https://westsidetoastmasters.com/resources/roberts_rules/chap12.html Approval Voting - https://en.wikipedia.org/wiki/Approval_voting Range Voting - https://en.wikipedia.org/wiki/Score_voting

Or some combination, such as a NomCom with an AC Approval ballot. I haven't done a lot of research on this, but I'm particularly intrigued by Range/Score voting as sort of a middle ground between the the traditional voting method W3C voting method, STV, and approval voting.

An alternative would be to accept factionalism as a fact of life, continue to use Meek STV as a way to select decision making bodies, but adopt majority voting as the decision making principle in the TAG, AB, AC, etc. That seems unlikely to produce credible industry standards that the losing side would implement.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/w3c/w3process/issues/484#issuecomment-749684243, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAD3Y6NO27NZ42KXRVETO63SWDLLVANCNFSM4VESXIAQ .

[chaals]

@cwilso wrote:

but disagree on the effect - in fact, factions can frequently get TWO candidates elected (one per cycle), rather than one.

Which is to say that we agree on the effect. If you term "representation of diversity" as "factionalism", what happens is the half as many "factions" (the largest/best-represented groups) potentially get two candidates elected, as you note, rather than a more diverse spread.

I think it is a good idea to use "no other candidate", and I think it is a good idea to allow for equal ranking. Both of these are absent largely because of operational decisions not to implement them.

[frivoal]

STV is not influenced by whether a candidate aims to compromise and work collegially or not. It is influenced by what the voters actually want.

If a candidate professes an intent to serve the interests of one group at the exclusion of everybody else, they may be ranked 1st by that that group, and last by everybody else. If another candidate professes an intent to find compromise, they may be ranked second by everybody. STV will elect the first one over the second one, provided that the group they pander to is large enough to exceed the threshold (which isn't particularly high, especially if you run STV over many seats at once, as suggested in #486).

When electing people to something like a parliament, that is fine, because once elected, representatives with minority viewpoints will have to seek some amount of compromise to get anything done. If a faction with 10% of parliament is the only one to hold a particular view which everyone else opposes, and won't budge on anything, they'll just fail to achieve anything. The AB and TAG, on the other hand, work like teams, not like parliaments, and run essentially by consensus. A single uncompromising radical can block everything.

[bkardell]

Particularly the combination of STV and low turnout that makes the problems repeated raised troubling to me. I'm unsure how to deal with that, but I think that potential amplification issues which would be feathered out with larger turnout seem much worse.

[cwilso]

I would like to recommend that there be at least a trial experiment run of approval voting at the next election.

[michaelchampion]

"Approval Voting" is essentially the pre-2017 election system except that there's no limit on the number of people one can vote for, right? And winners are the top N approval-getters, where N is the number of open seats?

[dwsinger]

@michaelchampion That's my understanding. I think that there is also a subtle difference in voter instructions – "Please indicate which of the following candidates you would support/be-ok-with serving on the XXX."

I think any change of voting system is going to have to address:

• diversity, and voting blocks: the primary offered rationale for STV was to improve diversity. We can't be silent on this. We shouldn't have 'voting blocks' but we do have at least loose aggregations (e.g. by region), and the largest of those should not be allowed to take all the seats.
• strategic voting: does the system induce people to vote other than their actual intent, for strategic effect?
• failure modes: what sort of circumstances could rise to a result that's not the intuitively expected outcome?
• implementation problems: we do not have equal-rank implemented in STV, for example

For example, under Approval voting, if I really want X elected, it makes sense to list only X, as any other names listed increases their count and raises the possibility that they will outvote X, even if there are others I'd be perfectly happy to see serve.

I would tend to argue that the fact that it's hard to work out in your mind what the STV/Meeks cascade system does may be a good thing: just vote the candidates you would be happy to see serve, in your order of preference. You only need to intuitively understand the cascade if you are trying to vote strategically rather than simply stating your preference, and strategic voting distorts what voters express and hence also distorts the result. I'm not sure you need to understand the algorithm, just the principles that it embodies.

I think we also need a discussion of what we're trying to optimize for. Here are some examples:

• we want an XXX which maximizes the number of members who feel that those serving can or do represent their interests;
• we want an XXX that has a diverse range of skills and perspectives, so we understand the broadest range (this comes up for the TAG, for example, where there is a sense that we need people who understand i18n, security, accessibility, offline, etc.)
• we want an XXX where key parts of the ecosystem are represented (the usual cited part is user-agent creators, but we have also agonized for years over the poor representation of web-site creators and owners)
• we want people on the XXX who have experience with the organization and its operations (this is probably better handled as a nomination requirement, or by letting the voters de-select people who lack it, rather than by choice of voting system)

[michaelchampion]

I think we also need a discussion of what we're trying to optimize for. Here are some examples:

Very good point! iThis might be a separate issue, since tweaking the election system is unlikely to achieve all the relevant objectives but it will be hard to run an election experiment without success criteria.

The success criteria probably are different for the TAG, AB, and an eventual Board of Directors:

• The TAG needs people with technical expertise in the web platform who can make and defend technically sound AND principled decisions.

• The AB needs generalists who both understand technology and business/strategy/politics and can work with others make decisions most constituencies can live with.

• The Board needs people who can make hard decisions and stick with them even if some constituencies don't like them, e.g. to replace a well-connected but ineffective CEO, make strategically focused budget decisions rather than "spreading the peanut butter evenly")

• All need diversity on all sorts of dimensions to both make good decisions that will stick with W3C's diverse membership (see https://hbr.org/2016/11/why-diverse-teams-are-smarter and many similar articles a web search will return)

[bkardell]

If I could offer a very different take on this... After thinking about this for a really long time now, this conversation reminds me a little of the Douglas Adams quote from Hitchhiker's Guide...

“This planet has - or rather had - a problem, which was this: most of the people living on it were unhappy for pretty much of the time. Many solutions were suggested for this problem, but most of these were largely concerned with the movement of small green pieces of paper, which was odd because on the whole it wasn't the small green pieces of paper that were unhappy.”

I can't help but think we've become so hyper-focused on this discussion about how we count ballots and what that optimizes for that we've kind of lost sight of the fact that it wasn't the ballots that were unhappy in the first place.

There are practical bounds that prohibit achieving all of the desired qualities of the platonic ideal TAG or AB at the same time, so it is inevitably an optimization problem full of tradeoffs. What is optimal for any one election inevitably involves all sorts of complex factors, from considering the current state of the TAG, standards, technology and the web. This requires planning, outreach to manage and ensure we have good nominees, thoughtful discussion and concensus building.

In the past 7 years or so, at least, there has been lots of this going on informally - more like a caucus that helps vet potential candidates and encourages members to nominate qualified and willing people, and build consensus around them. The AB and TAG became more effective because of these discussions (pre-STV, and after a while, post too).

I'd like to suggest that we open a series of issues that address the larger process of selecting AB and TAG members, starting from nomination and leading all the way through ballot counting. We should "lean into" practices that have been shown to work. A nominination/election mechanism I think would be way more optimal is:

• A group of people who can look at elections ahead of time and help line up candidates with qualities we need who can run, and sometimes help them do so (finding someone to nominate them if necessary, or funding for them, sometimes).
• A group of people who can caucus and discuss the available candidates and why they believe a particular mix is optimal and convey this to the AC. Maybe that group can come to concesus on a single slate, or maybe there are a few with varying possibilities, each meeting some minimal bar of support.
• Recommendations are presented to ACs who can approval vote a slate.

• If you wanted to go a step further, perhaps slates could be ranked in terms of preference. Personally, I like the idea of an expressive vote, it's just that "Which individual do you prefer the most" isn't a question that does nothing for my desire to express what I think an optimal result is.

[michaelchampion]

I very much like Brian's suggestion above ... but sounds a bit like a NomCom, which I suspect will get resistance given previous CG and AB discussions. But maybe there is a creative middle ground between the current system (with informal caucusing to do some of the work NomComs do, but hindered by STV's unfriendliness to "slates") and a formal NomCom that recommends a slate of voters for an approval ballot.

Simply moving from STV to an approval ballot would be a first step, so ideally we could close this issue by adopting the suggestion from @cwilso in https://github.com/w3c/w3process/issues/484#issuecomment-759114344 and (unless serious problems emerge) move to the approval system in the future.

Another step would be to make the nominee recruiting process more transparent. Maybe there could be a CG that serves as the "caucus" Brian describes, where people could make their interest known, members could see which interested people they might want to nominate and/or support, and like-minded people could sort out whether to run in the upcoming election or the subsequent one. This would also give opportunities to consider whether additional people need to be recruited to offer more diversity in backgrounds or specialized skills needed at a particular point in W3C's evolution.

A further step would be to consider whether actual nominations should be transparent in real-time. Again there has been debate about this, but without a formal NomCom process, the "nomination caucus" CG would need to be able to see how actual nominations are going to really work well.

[dwsinger]

@bkardell Very much, yes. I made a mistake several years ago, when I suggested what organizations other than the IETF call a Nominating Committee; a group charged with making sure that there is a slate of excellent, skilled, experienced, balanced candidates. Not that controls whether other nominations are possible, just one that asks questions and tries to get good candidates on the ballot. Unfortunately the IETF NomCom is different (it presents a set of names to be up/down agreed-to as a block), and the suggestion got lost. But a non-profit I was board-chair of years ago did this; those termed-out and I think the previous year's termed-out people were 'expected' to serve, and they could rope-in others.

We don't have term limits, so something more like how the IETF forms the NomCom would probably be better. There are details, of course:

• does the NomCom get to see individual AC noms before the nom period closes (yes, I think, if the point is to balance the slate, and also so they don't waste time pursuing and nominating someone already nominated)
• do the NomCom candidates appear differently on the ballot (I think not; but perhaps the NomCom members and others are encouraged to say what they think is a good slate)

[cwilso]

So many things to unpack here.

I've shown resistance to the NomCom idea in the past, frankly, because it pushes the problem upstream - someone, somewhere, is "deciding who should be on the slate", and whether that is in a smoke-filled back room or not, it's definitely a bit obscure. At the same time, yes, the word "slate" more than the phrase "Nomination Committee" has given the impression they would do an IETF-style "vote yes or no on this block" slate.

That said, a primary frustation I have (and have clearly displayed :) ) with STV is that it promotes factionalism, and enforces diversity at the cost of moderates - and yes, at the expense of browser vendors too, as the past few elections have shown (as Microsoft, Google, and Mozilla candidates have all lost AB and TAG elections). Many might be happy with this outcome, as they perceive those browser vendors "have too much power" anyway - but I think it is a concerning disengagement of the very platform companies that should be engaged in the TAG and AB (lest they veer off in other directions), and I think STV encourages this diversity. (I use that word intentionally; I'm not implying this is a horrible thing at all, and I want to be clear that I think everyone elected by STV deserves their seat, but it is a second-order effect of STV that we had no clear indication of.)

This is why I suggested that the AB needed to add a few seats - and we did (https://github.com/w3c/w3process/issues/190). I still think this is of concern, and perhaps the TAG needs expansion as well. (Thus, https://github.com/w3c/w3process/issues/465.)

[mnot]

I've shown resistance to the NomCom idea in the past, frankly, because it pushes the problem upstream - someone, somewhere, is "deciding who should be on the slate", and whether that is in a smoke-filled back room or not, it's definitely a bit obscure. At the same time, yes, the word "slate" more than the phrase "Nomination Committee" has given the impression they would do an IETF-style "vote yes or no on this block" slate.

The IETF NOMCOM comes up with the slate; the confirming bodies can approve a slate, or send it back with comments.

That said, a primary frustation I have (and have clearly displayed :) ) with STV is that it promotes factionalism, and enforces diversity at the cost of moderates

citation needed

Seriously, I see a lot of unsupported assertions based on how people feel in this thread. Data, please.

[dwsinger]

I've shown resistance to the NomCom idea in the past, frankly, because it pushes the problem upstream - someone, somewhere, is "deciding who should be on the slate", and whether that is in a smoke-filled back room or not, it's definitely a bit obscure.

I have to push back. This is emphatically not what I am suggesting. I suggest we leave open direct nomination, and add to that a group who thinks it's their job to try to make sure there are

1. enough candidates to fill the seats
2. a balanced set of excellent candidates

There is no decision made in a smoke-filled room. You want Mordred on the ballot? Then directly nominate. I do not propose or want a NomCom that decides on the slate for simple up/down en masse approval. I do think that tasks that are "everyone's job" (thinking about who would be good, what skills may be needed in the upcoming period, and so on) too easily become "no-one's job", and that having a group think about this and ensure that we have an adequate slate (in both the numeric and capability sense) is good insurance.

[cwilso]

That said, a primary frustation I have (and have clearly displayed :) ) with STV is that it promotes factionalism, and enforces diversity at the cost of moderates

citation needed

Seriously, I see a lot of unsupported assertions based on how people feel in this thread. Data, please.

I'm not sure what data you want to see. The stated goal of using STV was to "promote diversity"; that is "avoid moderates" by another name. I can give anecdotal evidence from nearly every election we've run with STV in W3C - comparing candidates who won, when it's clear they had a "faction" while those who lost did not. There are of course other factors there - but people vote how they "feel", so those feelings seem to hold true. That is not, of course, hard data.

I could write up a page like https://www.uk-engage.org/2013/06/what-are-the-advantages-and-disadvantages-of-using-the-single-transferable-vote-stv-system/ applied to the W3C, but it's not far off as it is, and of course without loads of elections, it's hard to apply.

[cwilso]

I've shown resistance to the NomCom idea in the past, frankly, because it pushes the problem upstream - someone, somewhere, is "deciding who should be on the slate", and whether that is in a smoke-filled back room or not, it's definitely a bit obscure.

I have to push back. This is emphatically not what I am suggesting. I suggest we leave open direct nomination, and add to that a group who thinks it's their job to try to make sure there are

1. enough candidates to fill the seats
2. a balanced set of excellent candidates

There is no decision made in a smoke-filled room. You want Mordred on the ballot? Then directly nominate. I do not propose or want a NomCom that decides on the slate for simple up/down en masse approval. I do think that tasks that are "everyone's job" (thinking about who would be good, what skills may be needed in the upcoming period, and so on) too easily become "no-one's job", and that having a group think about this and ensure that we have an adequate slate (in both the numeric and capability sense) is good insurance.

That's great. That's just "a group who encourage people to run." That is NOT, however, what @mnot described in the IETF NomCom, and probably REALLY important for us to have a shared description of what we're talking about here.

[dwsinger]

I think I would want that Nominations can be directly made by/from the NomCom as well as by AC reps, and that the NomCom is also privileged to see all nominations as they occur. As I said in the comment above, this suggestion got lost by being confused with the IETF style (which to my eye is closer to being electoral college in effect, as they decide who'll get elected, and not a committee tasked with making nominations for an election).

The obvious and historical name for a committee that is tasked with ensuring that there are enough good nominees for an election is Nominating Committee, but we keep getting entangled with the IETF's different use of that name. Is there a better name we can use so as to distance from the IETF's electoral college?

[cwilso]

Who should compose that Nominating Committee? I'm concerned by giving the nomcom too many nominations directly; however, as you said, they will need to have insight into the nominations as they occur.

I don't mind using the name Nominating (or Nominations?) Committee - but we should avoid using the term "slate".

[dwsinger]

I suspect we would need to form the NomCom from a few ex-officio members, but otherwise by

• random selection from
• a set of volunteers drawn from
• the set of qualified potential volunteers
• excluding the set of disqualified people

Rather like the way that the IETF forms the NomCom. Getting the qualifications right would be hard (e.g. the IETF requires attendance at 3 of the most recent 5 meetings, because they are meeting based). Current chairs? Editors/ Who have served for at least N months? Who have chaired at least N meetings in the last M months? Or delivered at least N documents in the last M months? Members of the consortium who have served in the past on the body to be elected to (i.e. ex-members of the AB or TAG)?

[chaals]

@cwilso wrote

The stated goal of using STV was to "promote diversity"; that is "avoid moderates" by another name.

No, diversity is not "avoid moderates". It is (and was repeatedly explained as) an effort to avoid one "faction" (to use your terminology) being the only one represented, and has nothing at all to do with whether the candidates people prefer are those they trust to work toward consensus, or those they trust to fight to the bitter end on every little point.

Realities are more complex. My observation is that people will generally hold very tight to a couple of issues, like changing the way people get elected, and be very happy to work hard and find a compromise on others like whether there should be a quid pro quo in security policies as part of allowing browser vendors to add a black box that supports DRM systems in the user agent... or vice versa...

Several times, individuals who could be said to represent the "faction" of browser developers have not been elected. But I can't think of a time in the last decade when there haven't been at least one of the "four usual suspect" browsers represented on the AB, and their level of representation on the TAG over the last decade seems to be pretty substantial. I have not seen candidates elected in their place whom I would describe as "factional candidates who won't look for consensus", but I have seen a broader diversity of candidates.

(There are wide regional variations in the browser market. Chrome is clearly the market dominator, but by very different fractions in different regions. A number of browsers are more relevant in important regions than the other three of the "usual four", and if you lump them together the history of browser makers dominating the elections is even more pronounced).

There seems to be an issue here of who "should" be on the AB and TAG, or what "factions" "deserve" seats as if by right, in order to ensure those bodies represent W3C. I'll open a seperate issue for that, although I think it is integral to any Nomination Committee (understood as essentially promoting a slate we "expect" to be elected).

(I also note that the split election cycle actively serves to reduce diversity - hence #486.)

[nigelmegitt]

@chaals wrote

@cwilso wrote

The stated goal of using STV was to "promote diversity"; that is "avoid moderates" by another name.

No, diversity is not "avoid moderates". It is (and was repeatedly explained as) an effort to avoid one "faction" (to use your terminology) being the only one represented, and has nothing at all to do with whether the candidates people prefer are those they trust to work toward consensus, or those they trust to fight to the bitter end on every little point.

+1 to "promote diversity" != "avoid moderates" - I don't mean to raise the temperature but I want to call out that the assertion that they are equivalent itself seems somewhat offensive to me - @cwilso I very much doubt that was your intention.

A more useful lens here is that diversity is better representation of the people who make up the key constituents of the web. Consider a scenario in which the a group has a membership that can be considered diverse and representative. There's no particular reason to believe that those members would not be "moderate"; however I would expect them to bring more varied and relevant perspectives and experiences to the discussion leading up to any decision, thus improving the (subjective) quality of that decision.

If we're thinking about alternative voting schemes, we should be optimising in favour of such representation. Understanding the key constituents who should be represented is vital, of course.

[my views, not verified as also my organisation's]

[cwilso]

Realities are more complex.

Indeed, at least on this we agree.

Several times, individuals who could be said to represent the "faction" of browser developers have not been elected.

There is no such "faction" - or more to the point, that faction is tiny, and does not vote for the same candidates.

But I can't think of a time in the last decade when there haven't been at least one of the "four usual suspect" browsers represented on the AB, and their level of representation on the TAG over the last decade seems to be pretty substantial.

You are correct, there hasn't been a time in the last decade where there hasn't been at least one of the browser vendors on the AB and TAG. That is, however, irrelevant, as browser vendors don't have the same perspectives - and the browser vendors represent 100% of the users of web, as each user uses one of them.

I have not seen candidates elected in their place whom I would describe as "factional candidates who won't look for consensus", but I have seen a broader diversity of candidates.

"Who will [or won't] look for consensus" is irrelevant.

(There are wide regional variations in the browser market. Chrome is clearly the market dominator, but by very different fractions in different regions. A number of browsers are more relevant in important regions than the other three of the "usual four", and if you lump them together the history of browser makers dominating the elections is even more pronounced).

Indeed, and I wish they engaged themselves in the W3C more.

I'm not sure what point you are trying to make by saying "if you lump them together", but my point is still the same - lumping together all browser vendors (as if they will have similar perspectives) is foolish. Should we have additional perspectives, and more diversity? Sure. But when you do that at the cost of engaging those who are actually providing the platform to users, I would argue you're setting yourself up for irrelevance.

There seems to be an issue here of who "should" be on the AB and TAG, or what "factions" "deserve" seats as if by right, in order to ensure those bodies represent W3C. I'll open a seperate issue for that, although I think it is integral to any Nomination Committee (understood as essentially promoting a slate we "expect" to be elected).

Hah. Note there are very different concepts of NomComs and "slates" - cf https://github.com/w3c/w3process/issues/484#issuecomment-763864501. Again, browser vendors - and particularly engine implementers - are not a "faction" in the same sense. There's no "Chrome faction" -there's a singular Google vote.

As for "deserves seats as if by right" - no, that is not what I said, and certainly not "in order to ensure those bodies represent W3C". If you think an AB and TAG with between one and zero browser vendors on it is ideal, though - you've probably picked the right system.

[fantasai]

A small suggestion: take one of the seats from each election and elect it by rerunning the ballots through a Borda Count algorithm (assigning points based on position in the list) after discarding the STV-elected candidates. That has a decent chance of pulling in someone with strong second/third-place positions but not enough first-place positions.

[dwsinger]

Borda count wiki page

[cwilso]

I would not recommend making the voting system even more complex, personally.

[dwsinger]

I would note, having read the Wiki, that Borda has many of the characteristics of approval voting, and is a tabulation system choice; choice of tabulation system is not mandated by the process, so we could adopt it without a process change. It might be illustrative to re-run recent election tabulations with Borda rather than Meek.

[fantasai]

@dwsinger We could certainly re-run the elections using existing ballots if we're curious, but the Process requires Single Transferrable Vote while Borda Count, although it is also a form of Ranked Choice Voting, is not a Single Transferrable Vote system. So we would need a Process change to actually switch.

[michaelchampion]

Given that the precise tabulation system is not baked into the process, making an essentially editorial change s/Single Transferrable Vote / Ranked Choice Voting/ might be a good thing to do in Process 2021 to keep options such as https://github.com/w3c/w3process/issues/484#issuecomment-785187614 open.

[fantasai]

@michaelchampion That would not be an editorial change. It concretely affects the conformance of possible implementations.

[dwsinger]

I don't think that we considered in detail what was formally meant by "a Single Transferable Vote system" and what tabulation methods are considered (by those skilled in the art) to be STV and which are not. Conceptually, I think we might have intuited this to mean a method where the voters vote by ranking preference. It wouldn't surprise me to find mentions of tabulation methods in our discussions that are not formally STV methods, for example. Until Elika pointed it out, it had not struck me that Borda tabulation is not formally STV (as it doesn't "transfer" votes down the preference list, I assume). So I agree that we can't do what I suggested and simply adopt Borda. But maybe, as part of this can of worms, we can add the question "did we mean formally STV, or systems where the voters vote by Ranked Choice"?

[michaelchampion]

essentially editorial ;-) . I can't imagine that the AC people who voted on Process 2017 would have distinguished between "STV" and "Ranked Choice". Not to mention that they also approved language about "one vote per open seat" the current implementaton does not conform with.

The challenge is to find a way forward, and Dave's idea in https://github.com/w3c/w3process/issues/484#issuecomment-785187614 seems worth pursuing.

[nigelmegitt]

I get nervous reading post-facto interpretations of "what we meant". We have to act on the basis that the votes cast were based on the text presented, not some other thing that we believe/hope/fear might have been what was really intended.

Which just means, if a change is desirable, write it down carefully, and present it again for another vote.

[dwsinger]

@nigelmegitt of course. It's maybe sad that we can't adopt Borda because it's not formally STV. If Michael wants to make the case to the AC that the phrase should be changed this year, I won't get in the way, but it would be described as a change and included in the vote.

[jeffjaffe]

I get nervous reading post-facto interpretations of "what we meant". We have to act on the basis that the votes cast were based on the text presented, not some other thing that we believe/hope/fear might have been what was really intended.

Exactly. The only reasonable interpretation of "what we meant" is "the thing we had done as an experiment in the previous couple of years".

[fantasai]

The AB has asked W3M to run an approval vote experiment during the upcoming AB election and will likely do so in the upcoming TAG election as well.

I've opened a separate issue on questions we might like to ask about the results; please add any suggestions there. https://github.com/w3c/w3process/issues/521