[EM] Summed-Ranks (SR)

[Michael Ossipoff]

Some years ago, I proposed a Borda version that meets FBC, but little else.



It seems to me that I called it Summed-Ranks (SR).



Summed-Ranks (SR):



Unlimited rankings, with equal ranking allowed.



The winner is the candidate with fewest candidates ranked over hilm,
as summed over all of the ballots.



[end of SR definition]



SR meets FBC, and probably all the criteria met by Approval, except
for Later-No-Help (LNHe) and 0-info LNHe, and IIAC.



It surely meets participation, Mono-Add-Top, and Mono-Add-Unique-Top,
because it's a point system.



For the same reason, it probably meets Consistency too.



If it's used with power-truncation, then it meets LNHe (not just 0-info LNHe).



Then, it seems to meet every critrerion that Approval meets, except for IIAC.



Power truncation:



On a ballot, all of the bottom-voted candidates are counted as having
everyone (including eachother) ranked over them.



A ballot votes a candidate at bottom if it votes someone over hir, and
doesn't vote hir over anyone.



(Power truncation is what makes Symmetrical ICT symmetrical).



When I refer to SR, I'll be referring to SR with power truncation. I
make power truncation part of the definition of SR.



I've been saying that using or proposing a rank method can't be
justified unless it meets CD and automatically avoids the chicken
dilemma. I still believe that.



SR doesn't qualify in that regard. But it's definition is so brief and
simple, and its count is so easy, that it doesn't have the full usual
costs of a rank method.



People like rank-balloting, rank methods.



In my limited polling experience, people seem to do better with
rankings than with rating. They seem to better vote in their interest
with rankings, in many of those polls.



(But FBC-failing rank methods wouldn't be nearly as good as Approval
or Score, for official public elecions, with our elecorate).



If I were conducting a poll, and the number of alternatives were great
enough to make Condorcet, including ICT, unfeasible, maybe sometimes
I'd use Score, and maybe sometimes I'd use SR. Asking people to rank
the alternatives is asking them less, compared to asking them to rate
the alternatives.



For example, I conducted a household poll on voting systems. The
alternatives were Approval, Score, Approval with optional delegation,
Score with optional delegatioin, and MJ.



Because there were only 4 alternatives that had a chance, and only 3
voters, I used rank-balloting, for counting by Symmetrical ICT. As it
happened, there was no equal ranking, and there was a CW. And so any
Condorcet method would have given the same answer. Of course,
typically that will be so when people rank sincerely, as they would in
a home-poll like this, and as they would not in an official public
election, given the electorate's beliefs and assumptions.



I didn't turn these voters against MJ. They've rejected
wordily-defined methods that I've liked. They didn't need me to tell
them that they didn't like MJ's definition.



Score was the wiinner. It beat every other alternative by at least 2
to 1. Everything beat MJ by 3 to 0.



Score was the Condorcet winner.



MJ was the Condorcet loser.



MJ scored below Approval and Score in the 0-1 Score poll on voting
systems, conducted at http://minguo.info



The only thing that did worse than MJ in that poll was traditional
unimproved Condorcet. (Improved Condorcet wasn't included in that
poll, because it isn't proposed or advocated for official public
elections.)



Anyway, I mention this here because ranking is easier than rating--is
less to ask of voters. When you actually conduct a poll, you know that
it's easier to ask people to mark their 1st choice, 2nd choice, etc.,
than to ask them to rate the alternatives sincerely.



Of course you could use SR as Approval, by using only two rank levels,
top and bottom.



And you could use Approval or SR as Score, probabilitically.



But there's no denying that strategic use of fractional ratings would
be easier in Score than in SR. For that reason, I'd prefer Score, when
strategy is a consideration. When no strategy is expected, and people
are expected to vote sincerely, then it's easier to ask them to rank
than to rate.



SR's method-assigned ratings, for the rank levels, might not be what
voters would choose if they rated sincerely, but, as i said, it's
easier to rank than to rate, and sometimes, often or usuallly, you'd
prefer not to ask people to rate, especially if the vote isn't on
something important, and especially if it's on a matter for which
enthusiasm or favoriteness is what really matters.



As I've said, for many small-group polling decisions, I'd prefer Score
balloting, electing the alternative whose lowest rating is the
highest.



I feel that that would be feasible for official public elections, in
an ideal society with completely honest voters who are caring and
aren't unduly selfish.



For voting on a movie, and maybe similar choices too, I'd likely use
Score or, more likely, SR.



Because SR has the same criterion-compliances as Approval and Score
(except for IIAC), SR is competitive with them for official public
elections (unless IIAC-failure defeats it), if its greater count-labor
isn't a fraud-security problem.  Obviously Score would be better,
because of its easy fractional ratings, whilch can be chosen to
reflect uncerttainty, or used for chicken-dilemma
defection-deterrence.



For official public elections, I'd prefer Approval or Score. But SR
would be ok if secure counting is feasible, and if people would prefer
it.



When you mention rank-balloting, people usually assume you mean Borda.
I've experienced that, and it was reported regarding the poll on
political parties, conducted within the Occupy Movement.



So who knows, SR might be a popular choice, for official public elections.



Mike Ossipoff