[EM] High5 voting (~ Smith//Approval on a reduced set)

[Rob Lanphier]

Hi Ted,

Thanks for the detailed writeup and comparison to Vote321.  I think "High5"
is a clever name, and I appreciate that it's not named after a dead white
guy (or maybe, perhaps, a living white guy).  I'm admittedly not a big fan
of the system, though (at least, not yet).

It seems to me that the practical difference in outcomes between High5 and
STAR would be so trivial as to be negligible.  It's my understanding that
STAR and pretty much all of the Condorcet methods VERY RARELY show
differences between them in simulations, and certainly, any contrived
scenario where there's a difference between Condorcet methods and STAR with
similar rankings/ratings feels contrived and highly unlikely.  My hunch is
that High5's performance would be imperceptibly different than STAR (and
most Condorcet methods), at a cost of much greater complexity and less
mainstream political attention than STAR (if one can count a ballot measure
in a small Oregon city as "mainstream").

In a "High5" election, I'm also not sure that any more than two or three of
the frontrunners would get mainstream attention (even given the name and
mechanics of the method).  My hunch is that mainstream electoral coverage
would do whatever it could to simplify all races down to the fewest number
possible.

Here's a different pair of systems which might achieve your stated goals
with High5: perhaps select five candidates to be selected via a "unified
primary"[1] (or rather, an approval-based primary).  Then the top five
would be running against one another for a few weeks/months before a
general election, without having to also compete with the crackpots who had
a successful signature campaign and managed to get on the primary ballot.
Use a Condorcet method (or a really good single-winner method) for the
general election round (as in High5's final tabulation).  This would put
five candidates "on the debate stage", so to speak.

There's a lot more that we could discuss about the value of two sequential
elections as is customary in the United States.  In short (if I understand
your original "High5" proposal correctly), it seems a mistake to try
consolidating the primary and the general election.   I think I'm going to
start a separate thread for the topic of primary+general elections versus
consolidated general elections.

Rob
[1]  The "unified primary" is basically the approval-based system used in
St. Louis:
https://electowiki.org/wiki/Unified_primary


On Fri, Dec 15, 2023 at 12:28 PM Ted Stern <dodecatheon at gmail.com> wrote:

> I've been mulling over ways to get a Smith/Approval method into a somewhat
> practical form, and have taken some cues from Jameson Quinn's Vote-321.
>
> Vote-321 is a pretty method, with a lot of strategy resistance, and if it
> were among the options for a new voting system (with no Condorcet method
> available), I would choose it immediately.
>
> However, I have a few issues with it: vulnerability to cloning; lack of
> expression; and what I consider a too-small provisional subset. The latter
> is more of a psychological / media problem, and what I mean by "too-small"
> is that if only the top three first-place candidates make it past the first
> pass, public attention could be excessively focused on the front-runners at
> the cost of addressing issues raised by less popular candidates. But the
> too-small subset is also what enables Vote-321's vulnerability to cloning.
> By including at least one or two more candidates in the first pass of
> candidate reduction, cloning risk is reduced.
>
> Following is what I call the *High 5 *method for three or more candidates:
>
> *Ballot Expression:*
> I prefer a 6 slot ranked ballot, equal-ranking and gaps allowed, with the
> rank/tiers named as follows:
>
>
> Tier Name
>
> Approval Status
>
> Description
>
> A
>
> Approved
>
> Most Preferred / Best / Favorite
>
> B
>
> Approved
>
> Good
>
> C
>
> Approved
>
> OK / Acceptable
>
> D
>
> Disapproved
>
> Not Preferred, but would be in their coalition (i.e. Compromise)
>
> E
>
> Disapproved
>
> Mostly Unacceptable but Lesser Evil
>
> Reject
>
> Dispapproved
>
> Completely unacceptable
>
> Summarized, there are 3 approved ranks (Most Preferred, Good, OK), 2
> disapproved ranks (compromise, lesser of two evils), and Reject. Blank
> ballots are counted as rejection.
>
> *Tabulation:*
>
>    - Total most-preferred votes per candidate (i.e. "A" votes).
>    - Approval / Disapproval totals per candidate
>    - Pairwise preference array
>    - Optional:
>       - Tied-Approval pairwise
>       - Tied-Disapproval (above reject) pairwise
>       - Approved vs Disapproved/Reject pairwise
>
> *First-pass subset:*
>
>    - *Top 5 candidates by most-preferred votes*
>
> *Procedure:*
>
>    - Of the top 5 most-preferred candidates are found, drop the
>    least-approved candidate.
>    - Among the remaining candidates, use the pairwise preference array to
>    find the Smith Set
>    - If more than one candidate is in the Smith Set, pick the most
>    approved member of the set as the winner.
>
> If you start with 3 candidates, this method reduces to top-two approval.
>
> Starting with 4 candidates, this method reduces to sorting the candidates
> by Approval and doing a top-three tournament: the winner is the pairwise
> winner of A1 versus (the pairwise winner of A2 versus A3).
>
> For 5 or more candidates, the method has a very high probability of
> finding the CW (if one exists) among the top five favorites, while falling
> back to approval in the event of a cycle among the four most-approved of
> those top five.
>
> In a "jungle-primary" type of situation (though no primary is necessary),
> media attention would be given to at least the top 5 candidates instead of
> just the top two, ensuring attention to a range of viewpoints. And in the
> event of a large number of candidates, it would not be necessary to
> tabulate pairwise preferences for more than 7 or 8 top-first-ranked
> candidates as determined by pre-election polling, reducing tabulation
> complexity while retaining summability.
>
> Using most-preferred votes for the first pass has a slight anti-cloning
> effect, as a preference ballot would lead to a tendency toward vote
> splitting if one faction has too many candidates. The anti-cloning pressure
> is greater than in Vote-321 because of the larger range of expression.
>
> Finally, the Smith//Approval method, when combined with the explicit
> approval cutoff ballot, allows enough strategy to reduce burial incentive.
>
> I'm calling this "High 5" voting because it's descriptive of both the
> ranking method and the top-five most preferred first-level truncation, and
> it's easier to remember than Smith//Approval.
>
> Thoughts?
>
> ----
> 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/20231217/92772914/attachment-0001.htm>