[EM] Strategy-free criterion

[robert bristow-johnson]

> On 06/13/2024 9:16 PM EDT Closed Limelike Curves <closed.limelike.curves at gmail.com> wrote:
> 
...
> Schulze generalizes Condorcet very simply by allowing indirect beats to be counted as well as direct ones. The definition is actually very intuitive and easy to describe in words:
> 1. A candidate has a direct win over another candidate if the first candidate is ranked higher than the second by a majority of voters.
> 2. An indirect win or chain is a series of direct wins going from one candidate to another.
> 3. The strength of a direct win is equal to the number of votes for the candidate preferred by a majority of voters.

That's WV not margins.

> 4. The strength of an indirect is equal to the strength of the weakest link in that chain (c.f. a chain is only as strong as its weakest link).
> 5. One candidate defeats another if the first candidate has a stronger chain going to the second candidate than vice-versa.

I do know and I did know how Schulze works.  But, conceptually, even the concepts above do not get encoded concisely into law with words.

> However, there's some real problems here. The biggest is that legislators and voters are not competent or intelligent enough to follow simple instructions.

For voters, the instructions are this: Rank the candidates in order from your most-preferred to least-preferred candidate.  Equal rankings are okay (this is not IRV).  Every candidate not ranked is considered ranked lower than any candidate who is ranked.  That's it.

For legislators, it ain't their job to "follow instructions".  It's their responsibility to closely examine and thoroughly understand the legislation that they are considering passing into law.

> I give a 30% chance that, even if you hand them a prewritten bill, the legislators will muck around with it and accidentally implement Schulze(margins), i.e. turkey-winning Schulze; or Schulze(pairwise opposition), i.e. Schulze with plurality failure.

You see, even though there are special interest groups that *do* get away with pre-writing legislation that gets rubber-stamped by a legislature, I would certainly say that the legislators (and the Office of Legislative Counsel) are not doing their job if they don't thoroughly examine and edit such legislation.  I sure as fuk don't want FairVote nor RCVRC writing legislation that gets passed verbatim into law.

> Second, even if they didn't, I don't trust them enough to expect whatever software lets them calculate the results will be thoroughly tested.

Uhm, in terms of tallying, I'm pretty sure that the Dominion software is well-tested.

One thing we do as election workers, on a day before the election (essentially the first day that we process mail-in ballots before the election) we test out the tabulator (and we practice) by starting up the election that it was pre-configured for, zeroing all the tallies, and processing about a couple dozen (maybe 30) ballots that *we* marked up and counted by hand.  Then we run them into the tabulator, then *close* the election day, which causes it to spit out the tallies on a strip of paper.  We, the Ward Clerk, the Inspectors of Election, and the City Clerk must all be satisfied that the tabulator counted those 25 or 30 ballots correctly.

> "But there's plenty of free, open-source, off-the-shelf implementations of Schulze! It can't be that hard to just pick one and—" is exactly what I used to say about IRV.

In 2009, with the AccuVote machines we had, they *did* use free software that I think the city got from Cambridge MA (who have been doing STV since the 1940s).  It was this "ChoicePlus Pro -- Version 2.3.2 (c) 1993-2006 Voting Solutions" and someone had to write a piece of glue code that would take the matrix of filled-in ovals and turn it into something the ChoicePlus Pro could read.

But governments don't want fuss and they want to rely on reputable and reliable manufacturers and we think that Dominion is such.  Which is why they had to sue the fuck outa Faux News for besmirching their reputation.

In their "Democracy Suite EMS Results Tally & Reporting User Guide", in the chapter for RCV, they don't have an option for any Condorcet yet.  Just IRV and STV (WIGM) and something called "Points IRV" and options.  But I have no doubt that if a state passed some Condorcet into law, that they would have any problem adding that to their list of options.

> Finally, even if the legislators get it got it right, I don't really trust voters to use it correctly.

It's the same damn ballot as IRV.  And the meaning of the ballot is the same:  Who's your favorite candidate?  Mark them #1.  Now, imagine that your favorite candidate was not running at all, from the group of remaining candidates, then who would be your choice?  Rank them #2.  etc.

If it's not a Score ballot, the meaning of the Ordinal ballot is the same whether it's IRV, Condorcet, Bucklin.  (Borda is a little iffy.)

> In particular, I'm starting to doubt that voters have the intelligence or understanding needed to play good strategies that elect the Condorcet winner in Schulze or RP, because they're not immediately obvious.

I agree.  That's one reason that I am not particularly worried about MO's (or yours, if I recall correctly) alarmism about Condorcet-Plurality devolving immediately into strategic voting (I assume burial) if that were adopted.  I don't think it would, even though I understand some scenarios in which a strategy could have worked (I think the Alaska special election in 2022 is a good example) to a candidate's benefit.

> Maybe they can work it out if you make the strategy blatantly obvious with implicit approval, but I doubt they could otherwise. I'll send an email with more details soon.

Looking forward to it.  I thought *both* you and MO were slamming my suggestion (and effort, I actually *did* get a bill written as such) to adopt Condorcet-Plurality.  I never disputed that, if the polls indicated that the Plurality winner was different than the Condorcet winner (say, the Alaska 2002 special election), that a conceivable strategy would be for supporters of the plurality winner to bury the Condorcet winner and throw the election into a cycle.  But it would take a concerted effort of tens of thousands of voters to mark their ballots insincerely (like not rank Begich, even though they like Begich better than Palin).  And it could conceivably backfire and cause the election of their most hated candidate.

I am not wedded to Condorcet-Plurality.  I think maybe Condorcet-TTR might be okay.  Maybe even Condorcet-Bucklin.  But clear legislative language that says what it means and means what it says is far easier to sell to legislators than something that is cryptic and the have to trust the lobbyists who wrote the legislation that it doesn't have some sneaky flaw that makes them regret passing the bill.  That's why I am promoting two-method Condorcet.  Then, at least, the policy makers (and citizenry) can read directly the intent of the legislation and not suspect that there is some sneaky clause that advantages a particular interest group.

--

r b-j . _ . _ . _ . _ rbj at audioimagination.com

"Imagination is more important than knowledge."

.
.
.