Experiment to disprove FailVote antiApproval arguments

From @parker_friedland on the old GoogleGroup.

  1. In a survey, ask people to honestly rate every candidate (including primary candidates) in three real world competitive elections (such as the 2008, 2012, and 2016 US presidential elections) from 0 to 5 stars on three separate pieces of paper (one piece of paper for each election and have them to do this in pen so they cannot erase).
  2. Ask everybody that participated in the survey to participate in a study to simulate the first of the three elections under some alternative voting method.
  3. Tell them that each of them will win 20 cents times the number of stars they rated the winning candidate.
  4. Randomly chose some percentage of the people participating in the study to participate in an anonymous poll about how they would vote in that election under the alternative voting method.
  5. Reveal those poll results.
  6. Repeat steps 4 and 5 two more times (but remember to conduct each poll one at a time so each poll has the ability to affect the next poll).
  7. Encourage the people participating in the study (but do not require them) to reveal the paper that has their honest ratings on them to other people in the study, as well as the ability to discuss strategies with people that have opinions that are similar to their own. The people participating in the study will have an hour to do this.
  8. Have everybody vote anonymously under the alternative voting method.
  9. Give everybody 20 cents times the number of stars they rated the winning candidate.
  10. Repeat steps 2 through 9 under the second election (so that if they feel that their strategy didn’t work, they will have the opportunity to try again with a new strategy).
  11. Repeat steps 2 through 9 under the final and most recent election. But now that the voters have gotten the hang of this, use 2 dollars (or more if you can afford it) instead of 20 cents so they have a much greater stake in the election.

I would like to propose a few tweaks of my own, though.

  • For step 1, add some background info on the third parties so that people do not just leave them blank. Also, perhaps require that every ballot include at least one “5” and at least one “0”?
  • Step 3 should probably be done after the first poll of step 4.
  • How does 9 work anonymously? Seems to take a lot of time.
  • For 10 it might be good to double the stakes… unless people will catch on and realize the third round is what “really matters”.
  • For the first and second election, we should probably manipulate it further so some people just vote honestly.

What do you think?

Wait, two more things:

  • One hour to debate may be a little too long, especially multiplied by three (or four, if we decide so). If this is an in-state thing, we should use Senate candidates if contested as well as presidential candidates.
  • Showing your star surveys outright should never be allowed. In real elections, people lie to pollsters. People can share their views through debate, but if they just straight-up reveal their utility function, it seems like a 50%+1 could force their 4-5 star candidate to win.
  • More information and more candidates should actually be added gradually. For example, 2008 uses just 4 candidates, but maybe 2012 includes 6, and 2016 adds a whole bunch of candidates so people are encouraged to consider third parties and independents (they “grow”).

If you’re willing to twist their arm that way, it might be worth it to try partial voting: https://www.docdroid.net/sQKg6nD/emerson-the-politics-of-consensus.pdf#page=36

Frankly, the arguments FairVote has against Approval fall apart under more high-stakes/with-polling scenarios. It’s only when voters don’t really care about an outcome, like in a college student body election (voted in those, can tell you that I myself bullet voted in an Approval Voting election where all but one of the candidates were bland) that you’d expect mass bullet voting and lack of polling strategy. But this is why I think lobbying for Score Voting in at least one city could help disprove the bullet voting argument, leading to Approval at the statewide level.

Of course, I would expect that out of the three levels of government (city/county, state, and federal), city elections are the most likely to have voter apathy. If Fargo’s next election ends up with too many bullet votes, I blame apathy. (Of course, FailVote will spout precisely the opposite interpretation.)

But the whole point of actually running a real-world experiment like this is to provide practical evidence, rather than the theoretical mathematics that they like to taunt. Seeing as we are probably not going to have Approval/Score elections for highly contested seats in the near future, running mini-experiments like this is the only way to get the practical evidence that would really disprove FailVote.

Addendum:
When giving out the rewards, I think it should be $2 per star in the first two rounds, and maybe $4 per star on the final round. Competing for fractions of a dollar seems pointless, and will probably lead to FailVote complaining that the first two elections were too informal.

I don’t believe the public will accept some elaborate experimental data to learn that ranked choice voting is a scam.

I am sure it is already easy enough to explain that the ranked choice, count eliminate, count eliminate, count eliminate, fake ‘majority’ counting scheme is a tricky, complicated gimmick that put all the power into the hands of the officials. It simply is not real voting at all, it is just a shell-game. It really can never be verified unless every single voter who makes a request is provided with a complete copy of every single individual ballot.

Score voting on the other hand, where the votes are all simply added up, is real voting.

And ‘ranked choice’ is a cashroots psyop to demolish democracy, which is promoted by millions of rich foundation dollars. The ordinary people who are conned into promoting it are simply being exploited by billionaire, so-called ‘philanthropists’ who hate democracy.

I have a yellow vest and a foam plastic backed sign, tied to my shoulders with safe, breakable, black plastic electrician’s tape, which says:

Ranked Choice Voting
Is An Oligarch Scam

It’s worth pointing out that RCV, by FairVote’s own standards, picks the same winner as vote-for-one 95% of the time. The real estimate ought to be closer to 90% of the time, because FPTP vote-splitting sometimes picks the wrong winner.

Such proportions depend heavily on the political environment, though. Consider, for example, how the frequency of unopposed candidacies, or races with two candidates only, would affect the failure rate.

This experiment is not designed to show that IRV is a scam. It is designed to show that Approval Voting does not lead to bullet voting when voters care about the outcome. (By extension, Score Voting should be even less likely to lead to bullet voting.) It is intended to show “FairVote” that approval is not the work of the devil.

It is actually quite difficult to tell this to average citizens. The reasons are quite technical, and to many, IRV is voting reform.

Although I already knew this, now I feel inspired to start telling pro-IRV politicians to support Score instead.

The problem with this sign is that in the mind of the average irrational human, this will taint all voting reforms. A better idea is to completely ignore IRV and promote the benefits of Simple Score Voting. If you do happen to get a “what about IRV” question, use the unverifiability, multi-round-near-tie-chaos, and “well it is just simpler” arguments.

Bug with Discourse: quoting your link popped up a warning asking if I was sure I wanted to duplicate the link…
Of course, Approval may also have a similar problem initially, although I would expect that as people get more used to thirds/indeps running, the “same as FPTP” proportion will go down (and cease to be meaningful as a number, as the entire slate of candidates would be different).

That is why, in my experiment model, I would start with fewer 3rds/inds and ratchet that up in later rounds of voting (perhaps including fake “Joe Centrist”, “Bob Far-Right”, “Ann Libertarian”, etc in the later elections).

1 Like

Asking the people breaking the system to fix it is vexing. As the intro to the Politics of Consensus book says, politicians hate unpredictability impacting them.

I’m partial to the “it’s more expensive” and “a candidate who everyone likes loses to the majority preference, condemning the 49% to nothing and starting the revenge cycle and fear you see in politics todays” arguments. Your average Joe Voter won’t care that the process can’t be verified, given that the experts endorse it and it has the most momentum.

Keep in mind that a lot of people oppose IRV because they find the algorithm difficult to understand, and therefore don’t trust it. (Source: every public hearing on this topic in Annapolis this year.) IRV being opaque doesn’t work in FairVote’s favor.

This is never going to work. Anything that threatens the ultimate rollout of STV is going to be vetoed by FairVote. The only thing that you might be able to get away with is Condorcet.