14 ms·

Alameda County admits tallying error in ranked choice voting

92 points by erehweb 4y ago

doktorhladnjak 4y ago
Looks like the SF Chronicle is now requiring a subscription to read any articles? I don’t live in the Bay Area to subscribe but this article looks interesting. Anyone have an archive link?
- pengaru 4y ago
  Loads fine here with noscript / js disabled fyi.
- doctorshady 4y ago
  https://archive.ph/w9Qpd https://archive.ph/w9Qpd
hodgesrm 4y ago
Wow. The complacency of election officials and many advocates for election reform is stunning. Ideas like ranked choices are worse than worthless if you don't implement them correctly. It undermines faith in the entire voting process.
im3w1l 4y ago
I wonder why there were so many incorrectly filled out ballots that minutiae of how they were processed was enough to swing the vote. Maybe they should take a look at how they were designed too.
- cavisne 4y ago
  And the fact that the vast majority were for one candidate. Looking at pictures of the example ballots it’s a confusing format optimized for voting machines, not humans. I really don’t understand the American obsession with voting machines, I believe there is always a manual count anyway and American election counting is extremely slow anyway.
  - toast0 4y ago
    > I really don’t understand the American obsession with voting machines, I believe there is always a manual count anyway and American election counting is extremely slow anyway. American elections usually have a whole ton of races on them. I've usually got about ten elected officials to weigh in on, and then five issues to give a yes or no to (down since I moved out of California). That would make manual counting all the races a major pain. AFAIK, ballots are machine counted, a sample is taken and hand counted, and as long as the sample is consistent with the machine count, and the machine count isn't too close, that's that. Unless a candidate wants to pay for a hand count of their race. But each state has its own rules, and races are generally handled at the county level, so there's room for a lot of variation. The slow parts of the counting is usually validating the ballots before opening them, which may not be legally allowed before election day and recently, many states allowing for ballots to arrive by mail after election day if they have a timely postmark (details about ballots that arrive by mail without a postmark vary, of course). Sometimes there's issues with transportation of ballots or polling places have to stay open for a long time for voters in line, and generally counting in a state won't begin until all polls are closed, which can be a coordination issue. Otherwise, counts are usually available night of, which is good enough for me? The certification process is slow, because it takes a lot of manual work (as discussed above).
    - SllX 4y ago
      It was 5 ballot sheets front and back in the most recent election in my part of California. It’s actually irritating. Most ballot measures shouldn’t be ballot measures. Pretty much anything that’s a constitutional amendment shouldn’t even be on the ballot, and the more elected officials there are, the less pressure there is on the more abjectly political parts of government to do a good job (legislatures, Governors/Mayors). We’re up to electing State Supreme Court justices now, and well, you can’t be held accountable for your judicial picks if you’re not the mook that put them in office, now can you? As much as I enjoy having a month or so to fill out my ballot at my leisure, I’m beginning to see a downside to that combined with voting machines, namely in that it allows the ballot to basically grow endlessly over time. The point of elected officials is these are the mooks hired to do the job of making laws, appointing their cabinet members, filling vacancies in the Courts and, prosecutors offices and law enforcement, and running the business of government in a politically accountable fashion so that we don’t have to.
rubyist5eva 4y ago
[flagged]
- cavisne 4y ago
  [flagged]
dataflow 4y ago
> As the results were being released, FairVote noticed issues specifically related to how to count a second- or third-choice vote if the voter did not select a first choice. Ballots with blank spots in ranked-choice races are considered “suspended.” > The Alameda County registrar explained that if a voter didn’t select a candidate as first choice, then the second choice should have been counted as the first choice in the first round. The same would occur in subsequent rounds moving lower choices up into the empty slot. Did they not have a unit test for this?! Now I'm curious how many unit tests they have, and what they cover. Shouldn't the software be public too? If voting machines are here to stay, the public ought to be able to see what code they're supposed to be running, no?
- toast0 4y ago
  > Shouldn't the software be public too? Eh? The inputs to the software need to be public and accurate, and then the public can run their own interpretations of the algorithms and yell when things aren't right. I'm assuming that's what FairVote, the organization mentioned in the article as alerting the county registrar, did.
- heavyset_go 4y ago
  > Shouldn't the software be public too? If voting machines are here to stay, the public ought to be able to see what code they're supposed to be running, no? You'd hope so, but governments just contract out to companies that build the machines and software, and those machines and software remain proprietary IP of the companies that built them.
aliljet 4y ago
Whoa. People in the region have been incensed about the registrar's slow count. The NAACP has called for recounts in a very very tight mayoral election. And there's been some serious confusion about how ranked choice voting works. In every case, the registrar has operated as untouchable and perfect. And somehow they got it so wrong a race was actually decided the wrong way in their original tally? Uh. What?
draw_down 4y ago
If the “certified” election results are potentially incorrect then people are right to lack faith in them. The problem is the lack of correctness, not the lack of faith. I feel like I see this a lot, this concern about second-order effects over concern about the problem itself. It feels like a way to draw attention away from the thing itself. Like, if a member of some particular group does something bad we start worrying about hypothetical repercussions against that group so we don’t have to talk about the act.
ojkelly 4y ago
My understanding is that calculating ranked choice isn’t the easiest thing to do. In Australia we have a national electoral commission (AEC) who runs the process, have perfected this over decades. So much so that they also run the elections for Registered Organisations (unions) [0]. The story is similar in our state elections where our electoral commissions are so we’ll regarded you can hire them to run elections for your company [1]. This is all to say, trust in the process and the fairness and accuracy of the outcome are paramount to democracy. And that we’ve been doing ranked choice for decades, it works—don’t write it off because there’s teething issues. I can see merits to the US approach of having thousands of effectively independent elections being run (for a county for example), and the Australian approach where it’s run nationally by a capable independent commission that holds itself to a very high standard. Stories like this one, where a county is trialling a different election process, are important to share and appreciate—not because they stuffed up, but because they fixed it. It’s one of those counterintuitive times where admitting failure/fault should build trust. Democracy is slow on purpose. Statistics get us results on election day, but they don’t change the outcome even when they’re wrong. To borrow some jargon, I think it’s more important democracy is observable than it is fast. Ranked choice voting is, to my knowledge, the most representative process to elect. No vote is wasted. You don’t have to game theory your choice of the lesser of two evils, you rank who you prefer. It captures your opinion, not the shadow groupthink choices. This is all to say, keep going it is worth it. [0] https://www.aec.gov.au/ieb/ https://www.aec.gov.au/ieb/ [1] https://www.vec.vic.gov.au/voting/types-of-elections/other-elections https://www.vec.vic.gov.au/voting/types-of-elections/other-e...
- dataflow 4y ago
  > You don’t have to game theory your choice of the lesser of two evils, you rank who you prefer You "have" [1] to game-theory your choice no matter what the voting system is. Ranked-choice is apparently believed [2] to be less susceptible to tactical voting than some other systems, though. (Not sure if this is quantitative or qualitative.) > IRV may incentivize forms of tactical voting (such as compromising) when voters have sufficient information about other voters' preferences, such as from accurate pre-election polling. FairVote mentions that monotonicity failure can lead to situations where "having more voters rank [a] candidate first, can cause [the candidate] to switch from being a winner to being a loser." [1] https://homepage.cs.uiowa.edu/~dwjones/voting/4980.2020/projects/zhang.pdf https://homepage.cs.uiowa.edu/~dwjones/voting/4980.2020/proj... [2] https://en.wikipedia.org/wiki/Instant-runoff_voting#Tactical_voting https://en.wikipedia.org/wiki/Instant-runoff_voting#Tactical...
uoaei 4y ago
I went deep into electoral systems after the 2016 election, and emerged a tentative ranked choice voting supporter for the momentum that FairVote was accruing at the time, but secretly I really considered approval voting an ideal next step in electoral reform: 1. Amending the law is relatively simple: replace any instance of only one with any when it comes to election laws regarding marking the ballot. 2. Amending the ballot is a minor change: just allow anyone to mark any (including none or all) of the bubbles representing each candidate on the ballot. The language on the ballot itself barely changes, the structure of the ballot can and would be identical to today's ballots. 3. Explaining the method to the public is dead simple: there's a pattern here, "instead of marking just one, now you can mark any of them" is a sentence that anyone can understand and implement. Explaining the calculation ("whoever gets the most votes, wins") is also dead simple. 4. Extending to multi-seat races is trivial: just take the top n vote-getters instead of the top 1. 5. It appeals to intuition: it does feel like "cast one lot per candidate you approve of" and aggregating that across a population is a fair way of distributing consensus. "Approval" means many things to many people but it is still a clear subjective threshold for each person. This also means that voters will consider the result harder to dispute on the grounds that it doesn't represent the people's will. 6. Tactical voting is straightforward and hard to do wrong: for what it's worth, tactical voting is always an option, and people are always going to use it. It's just how things work, at least for the foreseeable future. That being said, of all the systems, approval voting makes it the simplest to engage with and the least underhanded. Vote for people you want to promote in the election, don't vote for people you want to suppress in the election. I really don't see the downsides. Deep in the discussion boards of the internet where the election nerds lurk, they complain about this or that theoretical imperfection. But as for the options that have been suggested, this one stands out as the easiest to adopt societally, with the most bullet- and fool-proof implementation.
- lostdog 4y ago
  If you want A or B to win, but you really prefer A to B, then should you approve of B at all or not approve of them? This preference situation happens in every election, and approval voting has no answer to it. It's completely unclear what a voter should do to get the result they want.
godelski 4y ago
I'm a big advocate for cardinal style voting systems. One of the big reason is because the counting algorithms are trivial. They are a argmax(reduce_sum(votes)). That's trivially parallelizable (which is important for counting votes). Ordinal systems (ranked) frequently have to deal with many rounds which creates a lot of added complexity and errors are far easier. There's other reasons why I think these systems are superior even from a social choice aspect, but I think algorithmic complexity of vote tallying is an underrated aspect of voting systems that people need to pay far more attention to. This is a big aspect of "transparency" in voting. https://en.wikipedia.org/wiki/Cardinal_voting https://en.wikipedia.org/wiki/Cardinal_voting https://en.wikipedia.org/wiki/Ordinal_voting https://en.wikipedia.org/wiki/Ordinal_voting
datapolitical 4y ago
SF publishes the cast vote record…does alameda not? With the full CVR it should have been obvious they were fucking up the count, because anyone could replicate it.
- supreme_loquat 4y ago
  It's pretty common to not release the full CVR unfortunately. It makes the process much less transparent and also makes it harder to study how the different voting methods work in the real world.
andrewf 4y ago
How many people vote in Alameda county? Was building software to automate the counting really more efficient than doing it by hand with piles of ballots? In Australia RCV for a single electorate is done by hand. Each candidate can nominate a "scrutineer" to observe the process. Compared to doing it in a computer, it's like single stepping through the algorithm in a debugger. It's transparent as can be and a problem like this would be caught instantly. It scales fine! The process is roughly O(viable_candidates * voters). It's rare for more than 3 candidates to be viable, so in practice it reduces to O(voters) =~ O(electoral_workers) =~ O(population). As a "computer person" I'm a big fan of the manual process, but the implementation can be lacking: https://www.theguardian.com/world/2013/dec/06/missing-wa-senate-votes-lax-security https://www.theguardian.com/world/2013/dec/06/missing-wa-sen...
- duskwuff 4y ago
  > How many people vote in Alameda county? It's the seventh largest county in California (out of 58), with 1.6M residents. Hand-counting votes is out of the question.
  - andrewf 4y ago
    Looking through https://www.acgov.org/rovresults/rcv/248/ https://www.acgov.org/rovresults/rcv/248/ the largest RCV race was for Oakland mayor, with ~125k votes. Most races had <20k votes. An Australian lower house district averages ~105k votes, the largest has ~118k. Per https://www.acgov.org/rovresults/248/ https://www.acgov.org/rovresults/248/ the largest, non-RCV races (eg Attorney-General) had ~430k votes which still seems doable given the process is O(n). What I didn't account for in my response is the sheer number of positions voted for in a single election. An Australian voter typically votes for two things - House and Senate - and occasionally a single ballot measure on top. Australian Senate races are both more complicated (proportional representation) and cover much larger districts (New South Wales has ~5.4M voters). It can take weeks to declare the result, which is definitely a downside. https://www.aec.gov.au/voting/counting/files/senate-count-process.pdf https://www.aec.gov.au/voting/counting/files/senate-count-pr... Sources for the Australian numbers: https://peo.gov.au/understand-our-parliament/parliament-and-its-people/people-in-parliament/members-of-the-house-of-representatives/ https://peo.gov.au/understand-our-parliament/parliament-and-... https://en.wikipedia.org/wiki/Division_of_Durack https://en.wikipedia.org/wiki/Division_of_Durack https://www.abs.gov.au/statistics/people/population/national-state-and-territory-population/latest-release#states-and-territories https://www.abs.gov.au/statistics/people/population/national...
koolba 4y ago
This seems like a simple problem to solve if all the ballots are simply public. The voter isn’t relevant, only the full suite of ballot data. Then each campaign can run their own cross check to ensure the tally is performed per the actual rules. At the end of the day it’s the votes that matter, not the voter.
- specialist 4y ago
  I support releasing the ballots. Something like "time bounded privacy". So maybe a week or two after polls close. Or as part of the canvassing board's election certification process. To the best of my knowledge, throughout the USA all the election materials are destroyed in short order. Which nicely moot continued, ongoing debate. One old-timey justification was to make the certification final, giving the public some certainty in the results, rightly or wrongly, so they'd accept the outcome and move on. Well. We're way past that now. Perhaps greater (utter) transparency will calm everyone down again.
gorgoiler 4y ago
Ah, that most awkward of bugs — a logic error. Buried in the middle of the article: If a voter didn’t select a first choice, then the second choice should have been counted as the first choice in the first round. The same would occur in subsequent rounds moving lower choices up into the empty slot. Instead, the erroneous algorithm didn’t count any vote in a round if a space was blank. So effectively your choices are ranked based on the ordinals you chose, not on the absolute values of the ordinals. I wonder what happens to ballots that use non-standard symbols whose intent is clear. I would guess that a non-zero number of people write “A, B, C” or “first, second, third”. (Or even, in the context of this ranking-gaps bug, ballots that just have “second, third” on them.) Protocol implementations that are “liberal in what [they] accept” certainly make life overly complicated. Every new implementation of a DNS server or HTML renderer has to have all sorts of awful quirks because of a lack of strictness in historical (and now widely deployed) client implementations. One translation of this principle to ranked choice voting is that if that a voter does not correctly fill in all the boxes and use “1, 2, 3” then their vote just doesn’t count — it would make the tally algorithm much simpler — but that also goes against the principle that a tally algorithm should go out of its way to scry as much into the intent of a voter as possible, and reject as few ballots as it possibly can.
- qubex 4y ago
  You deserve an upvote just for your use of ‘scry’.
- rgblambda 4y ago
  >If a voter didn’t select a first choice, then the second choice should have been counted as the first choice in the first round. In my country, if a first preference isn't clearly marked on the ballot sheet, then the ballot is considered spoiled. This is a pencil and paper, counted by hand system. It looks like what counts as a defect in Alameda County is a perfectly working tallying machine in other parts of the world.
  - denton-scratch 4y ago
    I agree; if the instructions say you must select a first preference, then failing to do that should be a spoiled ballot. It's impossible, without interviewing the voter, to determine what their intention was. Having said that, ranked choice is harder for voters to understand, so it makes it easier for them to spoil their ballot. You can ameliorate that using electronic voting machines (which I disapprove of); and you can engage in various kinds of voter education. I'm from an FPTP country, and I wish we had ranked choice. But trying to read a clear intention to a ranking with no first choice is like reading tealeaves, and election officials shouldn't do it.
ShredKazoo 4y ago
I'm a fan of a group called the Center for Election Science which researches different voting methods. They have an article on their site which compares ranked choice with their preferred method, approval voting: https://electionscience.org/library/approval-voting-versus-irv/ https://electionscience.org/library/approval-voting-versus-i... Approval voting is much simpler to implement than ranked choice, and has theoretical advantages when it comes to voter satisfaction.
- specialist 4y ago
  Election Science is doing yeoman's work. And they just keep getting better. > RCV, however, cannot be counted in precincts or separate locations. It must be counted in a centralized area. Huh. The gold standard for election integrity is casting paper ballets at poll sites, to be tabulated onsite the moment the polls close. There are many of these very specific non-obvious minutiae which derail or moot most would-be reforms. Disallowing precinct-based tabulation is one such deal breaker. I studied election administration for ~10 years. I'm (super) embarrassed that this very important implementation detail of RCV had never occurred to me.
  - ShredKazoo 4y ago
    >Election Science is doing yeoman's work. And they just keep getting better. They just need a billionaire patron and they could dramatically transform American politics for the better. All those billionaires who give fat checks to political parties and candidates are doing it completely wrong. "Don't hate the player hate the game" -- but what if you could change the game?
    - stephencanon 4y ago
      The billionaires like the game—it gives them power!
      - ShredKazoo 4y ago
        Approval voting has been shown to increase voter satisfaction. I'll bet it would increase voter satisfaction for billionaires as well. They'd no longer be stuck with just a few bad candidates to choose between.
smitty1e 4y ago
Having served as an election officer for the last decade, one must say that RCV is to ballots what a complex dependency tree is to your IT project: something that may look swell under laboratory conditions, but Fortune help you in the real world, at scale (see Maricopa County for related woes). Arguments about election reform need to stress two points to gain traction with me: 1. Physical security of the (obviously tangible) ballot. 2. Physical security of the voter while marking the ballot. I first heard Scott Adams point out that any situation involving great reward (e.g. political power) and low risk (the number of convictions pertaining to election fraud is scant) is going to be rife with corruption. Give me: - Election Day. Make it a holiday. Staff and provision it to make a positive experience for all. - Minimal mail-in ballots. They are a huge physical security risk in both ways mentioned above. - Minimal early voting. "Election Season" is bollocks and security risk. There are reasons that half of Oregon wants to eject and join Idaho. And those reasons do not include satisfaction with their jacked up elections and the dodgy results thereof.
- denton-scratch 4y ago
  The UK has mail-in voting, which is necessarily early voting. I'm against it; you can't be sure that the voter wasn't coerced if they didn't vote in person. You need mail-in voting in specific cases, such as soldiers and diplomats serving overseas. But in general, I don't think it should be allowed to cast a ballot before the hustings are over. Voters can't be compelled to listen to political arguments; but it should be at least possible for a voter to know, for example, that their favoured candidate was convicted of rape the day before the election.
  - smitty1e 4y ago
    Firm concur. Then there is the guilt by association: the people I think shadiest are the ones pushing all of these "innovations". Many applications benefit from KISS, and voting is an important example.
- ojkelly 4y ago
  This is where I think Australia’s compulsory voting shows its strength. Every citizen has a responsibility to submit a ballot (either valid, invalid, or covered in dicks). And the state (or in our case an independent electoral commission) has a responsibility to collect a ballot from every voter. Which leads to making it as easy as possible for you to vote, with early voting, mail-in ballots and heaps of polling centres. The ballot problem is easier than you’d think though. Make effective corruption untenable, by involving a lot of people in the process. Everyone who can gain from the outcome should have equal presence as observers. And the tangibility of the ballot is crucial, it’s much harder to make 10,000 pieces of paper appear or disappear than to flip a bit.

92 points by erehweb 4y ago

Looks like the SF Chronicle is now requiring a subscription to read any articles? I don’t live in the Bay Area to subscribe but this article looks interesting. Anyone have an archive link?

pengaru 4y ago
Loads fine here with noscript / js disabled fyi.
doctorshady 4y ago
https://archive.ph/w9Qpd https://archive.ph/w9Qpd

hodgesrm 4y ago

Wow. The complacency of election officials and many advocates for election reform is stunning. Ideas like ranked choices are worse than worthless if you don't implement them correctly. It undermines faith in the entire voting process.

im3w1l 4y ago

I wonder why there were so many incorrectly filled out ballots that minutiae of how they were processed was enough to swing the vote. Maybe they should take a look at how they were designed too.

cavisne 4y ago
And the fact that the vast majority were for one candidate. Looking at pictures of the example ballots it’s a confusing format optimized for voting machines, not humans. I really don’t understand the American obsession with voting machines, I believe there is always a manual count anyway and American election counting is extremely slow anyway.
- toast0 4y ago
  > I really don’t understand the American obsession with voting machines, I believe there is always a manual count anyway and American election counting is extremely slow anyway. American elections usually have a whole ton of races on them. I've usually got about ten elected officials to weigh in on, and then five issues to give a yes or no to (down since I moved out of California). That would make manual counting all the races a major pain. AFAIK, ballots are machine counted, a sample is taken and hand counted, and as long as the sample is consistent with the machine count, and the machine count isn't too close, that's that. Unless a candidate wants to pay for a hand count of their race. But each state has its own rules, and races are generally handled at the county level, so there's room for a lot of variation. The slow parts of the counting is usually validating the ballots before opening them, which may not be legally allowed before election day and recently, many states allowing for ballots to arrive by mail after election day if they have a timely postmark (details about ballots that arrive by mail without a postmark vary, of course). Sometimes there's issues with transportation of ballots or polling places have to stay open for a long time for voters in line, and generally counting in a state won't begin until all polls are closed, which can be a coordination issue. Otherwise, counts are usually available night of, which is good enough for me? The certification process is slow, because it takes a lot of manual work (as discussed above).
  - SllX 4y ago
    It was 5 ballot sheets front and back in the most recent election in my part of California. It’s actually irritating. Most ballot measures shouldn’t be ballot measures. Pretty much anything that’s a constitutional amendment shouldn’t even be on the ballot, and the more elected officials there are, the less pressure there is on the more abjectly political parts of government to do a good job (legislatures, Governors/Mayors). We’re up to electing State Supreme Court justices now, and well, you can’t be held accountable for your judicial picks if you’re not the mook that put them in office, now can you? As much as I enjoy having a month or so to fill out my ballot at my leisure, I’m beginning to see a downside to that combined with voting machines, namely in that it allows the ballot to basically grow endlessly over time. The point of elected officials is these are the mooks hired to do the job of making laws, appointing their cabinet members, filling vacancies in the Courts and, prosecutors offices and law enforcement, and running the business of government in a politically accountable fashion so that we don’t have to.

rubyist5eva 4y ago

[flagged]

cavisne 4y ago
[flagged]

dataflow 4y ago

> As the results were being released, FairVote noticed issues specifically related to how to count a second- or third-choice vote if the voter did not select a first choice. Ballots with blank spots in ranked-choice races are considered “suspended.” > The Alameda County registrar explained that if a voter didn’t select a candidate as first choice, then the second choice should have been counted as the first choice in the first round. The same would occur in subsequent rounds moving lower choices up into the empty slot. Did they not have a unit test for this?! Now I'm curious how many unit tests they have, and what they cover. Shouldn't the software be public too? If voting machines are here to stay, the public ought to be able to see what code they're supposed to be running, no?

toast0 4y ago
> Shouldn't the software be public too? Eh? The inputs to the software need to be public and accurate, and then the public can run their own interpretations of the algorithms and yell when things aren't right. I'm assuming that's what FairVote, the organization mentioned in the article as alerting the county registrar, did.
heavyset_go 4y ago
> Shouldn't the software be public too? If voting machines are here to stay, the public ought to be able to see what code they're supposed to be running, no? You'd hope so, but governments just contract out to companies that build the machines and software, and those machines and software remain proprietary IP of the companies that built them.

aliljet 4y ago

Whoa. People in the region have been incensed about the registrar's slow count. The NAACP has called for recounts in a very very tight mayoral election. And there's been some serious confusion about how ranked choice voting works. In every case, the registrar has operated as untouchable and perfect. And somehow they got it so wrong a race was actually decided the wrong way in their original tally? Uh. What?

draw_down 4y ago

If the “certified” election results are potentially incorrect then people are right to lack faith in them. The problem is the lack of correctness, not the lack of faith. I feel like I see this a lot, this concern about second-order effects over concern about the problem itself. It feels like a way to draw attention away from the thing itself. Like, if a member of some particular group does something bad we start worrying about hypothetical repercussions against that group so we don’t have to talk about the act.

ojkelly 4y ago

My understanding is that calculating ranked choice isn’t the easiest thing to do. In Australia we have a national electoral commission (AEC) who runs the process, have perfected this over decades. So much so that they also run the elections for Registered Organisations (unions) [0]. The story is similar in our state elections where our electoral commissions are so we’ll regarded you can hire them to run elections for your company [1]. This is all to say, trust in the process and the fairness and accuracy of the outcome are paramount to democracy. And that we’ve been doing ranked choice for decades, it works—don’t write it off because there’s teething issues. I can see merits to the US approach of having thousands of effectively independent elections being run (for a county for example), and the Australian approach where it’s run nationally by a capable independent commission that holds itself to a very high standard. Stories like this one, where a county is trialling a different election process, are important to share and appreciate—not because they stuffed up, but because they fixed it. It’s one of those counterintuitive times where admitting failure/fault should build trust. Democracy is slow on purpose. Statistics get us results on election day, but they don’t change the outcome even when they’re wrong. To borrow some jargon, I think it’s more important democracy is observable than it is fast. Ranked choice voting is, to my knowledge, the most representative process to elect. No vote is wasted. You don’t have to game theory your choice of the lesser of two evils, you rank who you prefer. It captures your opinion, not the shadow groupthink choices. This is all to say, keep going it is worth it. [0] https://www.aec.gov.au/ieb/ https://www.aec.gov.au/ieb/ [1] https://www.vec.vic.gov.au/voting/types-of-elections/other-elections https://www.vec.vic.gov.au/voting/types-of-elections/other-e...

dataflow 4y ago
> You don’t have to game theory your choice of the lesser of two evils, you rank who you prefer You "have" [1] to game-theory your choice no matter what the voting system is. Ranked-choice is apparently believed [2] to be less susceptible to tactical voting than some other systems, though. (Not sure if this is quantitative or qualitative.) > IRV may incentivize forms of tactical voting (such as compromising) when voters have sufficient information about other voters' preferences, such as from accurate pre-election polling. FairVote mentions that monotonicity failure can lead to situations where "having more voters rank [a] candidate first, can cause [the candidate] to switch from being a winner to being a loser." [1] https://homepage.cs.uiowa.edu/~dwjones/voting/4980.2020/projects/zhang.pdf https://homepage.cs.uiowa.edu/~dwjones/voting/4980.2020/proj... [2] https://en.wikipedia.org/wiki/Instant-runoff_voting#Tactical_voting https://en.wikipedia.org/wiki/Instant-runoff_voting#Tactical...

uoaei 4y ago

I went deep into electoral systems after the 2016 election, and emerged a tentative ranked choice voting supporter for the momentum that FairVote was accruing at the time, but secretly I really considered approval voting an ideal next step in electoral reform: 1. Amending the law is relatively simple: replace any instance of only one with any when it comes to election laws regarding marking the ballot. 2. Amending the ballot is a minor change: just allow anyone to mark any (including none or all) of the bubbles representing each candidate on the ballot. The language on the ballot itself barely changes, the structure of the ballot can and would be identical to today's ballots. 3. Explaining the method to the public is dead simple: there's a pattern here, "instead of marking just one, now you can mark any of them" is a sentence that anyone can understand and implement. Explaining the calculation ("whoever gets the most votes, wins") is also dead simple. 4. Extending to multi-seat races is trivial: just take the top n vote-getters instead of the top 1. 5. It appeals to intuition: it does feel like "cast one lot per candidate you approve of" and aggregating that across a population is a fair way of distributing consensus. "Approval" means many things to many people but it is still a clear subjective threshold for each person. This also means that voters will consider the result harder to dispute on the grounds that it doesn't represent the people's will. 6. Tactical voting is straightforward and hard to do wrong: for what it's worth, tactical voting is always an option, and people are always going to use it. It's just how things work, at least for the foreseeable future. That being said, of all the systems, approval voting makes it the simplest to engage with and the least underhanded. Vote for people you want to promote in the election, don't vote for people you want to suppress in the election. I really don't see the downsides. Deep in the discussion boards of the internet where the election nerds lurk, they complain about this or that theoretical imperfection. But as for the options that have been suggested, this one stands out as the easiest to adopt societally, with the most bullet- and fool-proof implementation.

lostdog 4y ago
If you want A or B to win, but you really prefer A to B, then should you approve of B at all or not approve of them? This preference situation happens in every election, and approval voting has no answer to it. It's completely unclear what a voter should do to get the result they want.

godelski 4y ago

I'm a big advocate for cardinal style voting systems. One of the big reason is because the counting algorithms are trivial. They are a argmax(reduce_sum(votes)). That's trivially parallelizable (which is important for counting votes). Ordinal systems (ranked) frequently have to deal with many rounds which creates a lot of added complexity and errors are far easier. There's other reasons why I think these systems are superior even from a social choice aspect, but I think algorithmic complexity of vote tallying is an underrated aspect of voting systems that people need to pay far more attention to. This is a big aspect of "transparency" in voting. https://en.wikipedia.org/wiki/Cardinal_voting https://en.wikipedia.org/wiki/Cardinal_voting https://en.wikipedia.org/wiki/Ordinal_voting https://en.wikipedia.org/wiki/Ordinal_voting

datapolitical 4y ago

SF publishes the cast vote record…does alameda not? With the full CVR it should have been obvious they were fucking up the count, because anyone could replicate it.

supreme_loquat 4y ago
It's pretty common to not release the full CVR unfortunately. It makes the process much less transparent and also makes it harder to study how the different voting methods work in the real world.

andrewf 4y ago

How many people vote in Alameda county? Was building software to automate the counting really more efficient than doing it by hand with piles of ballots? In Australia RCV for a single electorate is done by hand. Each candidate can nominate a "scrutineer" to observe the process. Compared to doing it in a computer, it's like single stepping through the algorithm in a debugger. It's transparent as can be and a problem like this would be caught instantly. It scales fine! The process is roughly O(viable_candidates * voters). It's rare for more than 3 candidates to be viable, so in practice it reduces to O(voters) =~ O(electoral_workers) =~ O(population). As a "computer person" I'm a big fan of the manual process, but the implementation can be lacking: https://www.theguardian.com/world/2013/dec/06/missing-wa-senate-votes-lax-security https://www.theguardian.com/world/2013/dec/06/missing-wa-sen...

duskwuff 4y ago
> How many people vote in Alameda county? It's the seventh largest county in California (out of 58), with 1.6M residents. Hand-counting votes is out of the question.
- andrewf 4y ago
  Looking through https://www.acgov.org/rovresults/rcv/248/ https://www.acgov.org/rovresults/rcv/248/ the largest RCV race was for Oakland mayor, with ~125k votes. Most races had <20k votes. An Australian lower house district averages ~105k votes, the largest has ~118k. Per https://www.acgov.org/rovresults/248/ https://www.acgov.org/rovresults/248/ the largest, non-RCV races (eg Attorney-General) had ~430k votes which still seems doable given the process is O(n). What I didn't account for in my response is the sheer number of positions voted for in a single election. An Australian voter typically votes for two things - House and Senate - and occasionally a single ballot measure on top. Australian Senate races are both more complicated (proportional representation) and cover much larger districts (New South Wales has ~5.4M voters). It can take weeks to declare the result, which is definitely a downside. https://www.aec.gov.au/voting/counting/files/senate-count-process.pdf https://www.aec.gov.au/voting/counting/files/senate-count-pr... Sources for the Australian numbers: https://peo.gov.au/understand-our-parliament/parliament-and-its-people/people-in-parliament/members-of-the-house-of-representatives/ https://peo.gov.au/understand-our-parliament/parliament-and-... https://en.wikipedia.org/wiki/Division_of_Durack https://en.wikipedia.org/wiki/Division_of_Durack https://www.abs.gov.au/statistics/people/population/national-state-and-territory-population/latest-release#states-and-territories https://www.abs.gov.au/statistics/people/population/national...

koolba 4y ago

This seems like a simple problem to solve if all the ballots are simply public. The voter isn’t relevant, only the full suite of ballot data. Then each campaign can run their own cross check to ensure the tally is performed per the actual rules. At the end of the day it’s the votes that matter, not the voter.

specialist 4y ago
I support releasing the ballots. Something like "time bounded privacy". So maybe a week or two after polls close. Or as part of the canvassing board's election certification process. To the best of my knowledge, throughout the USA all the election materials are destroyed in short order. Which nicely moot continued, ongoing debate. One old-timey justification was to make the certification final, giving the public some certainty in the results, rightly or wrongly, so they'd accept the outcome and move on. Well. We're way past that now. Perhaps greater (utter) transparency will calm everyone down again.

gorgoiler 4y ago

Ah, that most awkward of bugs — a logic error. Buried in the middle of the article: If a voter didn’t select a first choice, then the second choice should have been counted as the first choice in the first round. The same would occur in subsequent rounds moving lower choices up into the empty slot. Instead, the erroneous algorithm didn’t count any vote in a round if a space was blank. So effectively your choices are ranked based on the ordinals you chose, not on the absolute values of the ordinals. I wonder what happens to ballots that use non-standard symbols whose intent is clear. I would guess that a non-zero number of people write “A, B, C” or “first, second, third”. (Or even, in the context of this ranking-gaps bug, ballots that just have “second, third” on them.) Protocol implementations that are “liberal in what [they] accept” certainly make life overly complicated. Every new implementation of a DNS server or HTML renderer has to have all sorts of awful quirks because of a lack of strictness in historical (and now widely deployed) client implementations. One translation of this principle to ranked choice voting is that if that a voter does not correctly fill in all the boxes and use “1, 2, 3” then their vote just doesn’t count — it would make the tally algorithm much simpler — but that also goes against the principle that a tally algorithm should go out of its way to scry as much into the intent of a voter as possible, and reject as few ballots as it possibly can.

qubex 4y ago
You deserve an upvote just for your use of ‘scry’.
rgblambda 4y ago
>If a voter didn’t select a first choice, then the second choice should have been counted as the first choice in the first round. In my country, if a first preference isn't clearly marked on the ballot sheet, then the ballot is considered spoiled. This is a pencil and paper, counted by hand system. It looks like what counts as a defect in Alameda County is a perfectly working tallying machine in other parts of the world.
- denton-scratch 4y ago
  I agree; if the instructions say you must select a first preference, then failing to do that should be a spoiled ballot. It's impossible, without interviewing the voter, to determine what their intention was. Having said that, ranked choice is harder for voters to understand, so it makes it easier for them to spoil their ballot. You can ameliorate that using electronic voting machines (which I disapprove of); and you can engage in various kinds of voter education. I'm from an FPTP country, and I wish we had ranked choice. But trying to read a clear intention to a ranking with no first choice is like reading tealeaves, and election officials shouldn't do it.

ShredKazoo 4y ago

I'm a fan of a group called the Center for Election Science which researches different voting methods. They have an article on their site which compares ranked choice with their preferred method, approval voting: https://electionscience.org/library/approval-voting-versus-irv/ https://electionscience.org/library/approval-voting-versus-i... Approval voting is much simpler to implement than ranked choice, and has theoretical advantages when it comes to voter satisfaction.

specialist 4y ago
Election Science is doing yeoman's work. And they just keep getting better. > RCV, however, cannot be counted in precincts or separate locations. It must be counted in a centralized area. Huh. The gold standard for election integrity is casting paper ballets at poll sites, to be tabulated onsite the moment the polls close. There are many of these very specific non-obvious minutiae which derail or moot most would-be reforms. Disallowing precinct-based tabulation is one such deal breaker. I studied election administration for ~10 years. I'm (super) embarrassed that this very important implementation detail of RCV had never occurred to me.
- ShredKazoo 4y ago
  >Election Science is doing yeoman's work. And they just keep getting better. They just need a billionaire patron and they could dramatically transform American politics for the better. All those billionaires who give fat checks to political parties and candidates are doing it completely wrong. "Don't hate the player hate the game" -- but what if you could change the game?
  - stephencanon 4y ago
    The billionaires like the game—it gives them power!
    - ShredKazoo 4y ago
      Approval voting has been shown to increase voter satisfaction. I'll bet it would increase voter satisfaction for billionaires as well. They'd no longer be stuck with just a few bad candidates to choose between.

smitty1e 4y ago

Having served as an election officer for the last decade, one must say that RCV is to ballots what a complex dependency tree is to your IT project: something that may look swell under laboratory conditions, but Fortune help you in the real world, at scale (see Maricopa County for related woes). Arguments about election reform need to stress two points to gain traction with me: 1. Physical security of the (obviously tangible) ballot. 2. Physical security of the voter while marking the ballot. I first heard Scott Adams point out that any situation involving great reward (e.g. political power) and low risk (the number of convictions pertaining to election fraud is scant) is going to be rife with corruption. Give me: - Election Day. Make it a holiday. Staff and provision it to make a positive experience for all. - Minimal mail-in ballots. They are a huge physical security risk in both ways mentioned above. - Minimal early voting. "Election Season" is bollocks and security risk. There are reasons that half of Oregon wants to eject and join Idaho. And those reasons do not include satisfaction with their jacked up elections and the dodgy results thereof.

denton-scratch 4y ago
The UK has mail-in voting, which is necessarily early voting. I'm against it; you can't be sure that the voter wasn't coerced if they didn't vote in person. You need mail-in voting in specific cases, such as soldiers and diplomats serving overseas. But in general, I don't think it should be allowed to cast a ballot before the hustings are over. Voters can't be compelled to listen to political arguments; but it should be at least possible for a voter to know, for example, that their favoured candidate was convicted of rape the day before the election.
- smitty1e 4y ago
  Firm concur. Then there is the guilt by association: the people I think shadiest are the ones pushing all of these "innovations". Many applications benefit from KISS, and voting is an important example.
ojkelly 4y ago
This is where I think Australia’s compulsory voting shows its strength. Every citizen has a responsibility to submit a ballot (either valid, invalid, or covered in dicks). And the state (or in our case an independent electoral commission) has a responsibility to collect a ballot from every voter. Which leads to making it as easy as possible for you to vote, with early voting, mail-in ballots and heaps of polling centres. The ballot problem is easier than you’d think though. Make effective corruption untenable, by involving a lot of people in the process. Everyone who can gain from the outcome should have equal presence as observers. And the tangibility of the ballot is crucial, it’s much harder to make 10,000 pieces of paper appear or disappear than to flip a bit.