[EM] "STAR voting..." paper by Wolk, Quinn & Ogren

[Toby Pereira]

 Thank you for posting this. A few things jump out to me.
They briefly discuss the advantages, but the 0 to 5 scoring does still seem fairly limited. They say that it conveys more information than a ranked ballot, but that's not something they've given an objective measure for. You can make a distinction between fewer candidates on a STAR ballot than on a ranked ballot, but you can indicate strength of preference. Amongst that, what's the measure of amount of information? Rob - you mention about whether rating is more intuitive than ranking. My general understanding is that for a small number of candidates, it's easier to rank, but for a large number it's easier to rate, as you just have to have a general opinion of each one rather than an exact pecking order of them all.
On pass/fail criteria, it's interesting that they consider the Equality criterion (where it should be possible to exactly cancel out a vote with an opposite one) to be one that should be passed completely, while generally saying that it's more about not failing any criterion badly. But also, some criteria are easier to pass than others. So by using this reasoning it's easy to hide behind quite a bad criterion failure. E.g., passing participation limits your options for voting methods, whereas I consider independence of clones to be a lot "cheaper". STAR fails independence of clones. In general, I'm not sure a method has to give much up to pass independence of clones.
And this also brings me to my next point - the simulation model. The voting method can change which candidates choose to stand, but I don't think this is modelled by them. STAR's clone failure could cause parties/factions to field two candidates in an attempt to  lock out the run-off.
I wouldn't call myself anti-STAR, but I'm not completely convinced by it as a method.
Toby

    On Friday, 29 December 2023 at 21:20:24 GMT, Rob Lanphier <roblan at gmail.com> wrote:  
 
 Hi everyone,
There's a paper regarding STAR voting that was recently made "open access":
https://doi.org/10.1007/s10602-022-09389-3
I've only skimmed the paper, but it seems sensible at first glance.  Much of what I got out of it is that most election-method criteria should be viewed as spectra rather than cold black-and-white binary options.  It also seems to do a pretty good job of pointing out many of the mathematical problems with RCV/IRV that lead to unpredictable outcomes in close-ish elections (which don't have to be that close to cause big problems).

I used to take a dogmatic black-and-white view of the Condorcet winner criterion, and bought into the FUD from FairVote and others about approval voting (including prolific members of this mailing list) about likely voter behavior in approval elections.  Now that I've seen plenty of credible simulations (from Yee et al) and spent some time explaining approval voting to random activists in California (and elsewhere), and now that "approval rating" has become solidly mainstream (thanks to FiveThirtyEight and others), it's a lot easier for me to imagine how mainstream voters would vote in big elections.  Since I have yet to see a credible simulation that shows a difference between approval and Condorcet methods, I've been pro-Approval.  I frequently give STAR advocates the benefit of the doubt, since (intuitively) I'm reasonably sure that STAR would also be indistinguishable from approval and Condorcet methods.
I'm not going to give the STAR folks a free pass, though, since they often spread FUD about rating being much more intuitive than ranking, citing some politically-naive pseudoscience.  Being anti-ranking is a very US-centric view that ignores the electorate in many countries throughout the world (and immigrants from those countries to the United States), who have likely been taught about ranked ballots in civics classes in their youth, and may have even voted in a few ranked-ballot elections and seen the results with their own eyes.  STV is very good at proportionality when selecting multiple seats.  Given the outsized influence that San Francisco has on tech (and thus, American culture), and given San Francisco's use of ranked ballots for nearly two decades, it is politically stupid for the STAR voting crowd to position themselves as anti-ranking.  People move around in the world (even people who typically vote for parties that would be considered fringe-y "third parties" in the United States).  Olympic gold medalists stand on "tier one" during the medal ceremony, not "tier five".  As aggravating as ranked ballots might be to STAR advocates, they need to rejoin the worldwide political reality to not sound like idiots to many people.

As far as I've read, this paper seems to avoid being the anti-ranking screed that I've grown tired of from some STAR advocates, and takes a more nuanced approach.  A huge problem with RCV/IRV (and for that matter, with traditional Borda) is the disallowance of tied rankings.  Ranking isn't the problem; trivial ballot spoilage is.  Summability is also a big problem, though "summability" is probably more subjective than it sounds (see the electowiki talk page discussion on the "Summability criterion" [1]).  The paper seems to be more "pro-tie-ranking", which seems politically smarter.  When I vote in complicated elections with too many candidates, I like being able to sort candidates into tiers ("yay", "okay, I guess", "nope"), and I'm guessing I'm not unique in that regard.
Anyway, please give the paper a read, and let the list know what you think!  I plan to give it a closer read when I get the chance, and I'd certainly appreciate having some idea from all y'all about how much attention the world should be giving this paper.
Rob
[1]: https://electowiki.org/wiki/Talk:Summability_criterion#Why_isn't_IRV_considered_summable_for_electing_the_dog_catcher?
----
Election-Methods mailing list - see https://electorama.com/em for list info
  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20240103/9e3abf8b/attachment-0001.htm>