Is Clone-Winner the same as the Independence from Clones Criterion (ICC)?
If so, I don't think MDDA3 could meet it. At least not in the way we've been
defining ICC lately. Not having seen the definition of Clone-Winner, I
probably shouldn't be posting about this.
In the version of ICC that James & I recognize, sincere voting by everyone
should ensure that adding a clone shouldn't change the matter of whether the
winner comes from the clone-set.
That's the preference version of ICC, for which a clone-set is defined in
terms of preferences rather than votes, and could contain as few as 1
candidate.
The disadvantage of defining ICC in terms of votes-only is that Approval and
Purality pass, which is contrary to what people intend with ICC. There are
ways that are used, with some votes-only criteria, to keep Plurality from
winning, but they're all pretty undesirable, as I've been discussing on EM.
Preference ICC is universally and uniformly applicable, and it makes no
mention of methods' rules, including their balloting rules. That can't be
said of the votes-only versions.
Anyway, say your preference ordering is: A>B>C. And say that {A,B} is a
clone-set. One way to vote sincerely would be to vote 1A, 2B.
Say B wins, by one vote, in the Approval phase of MDDA3.
Now we add C to the election. Say you like C better than A and B. Now, your
only sincere ballot is:
C>A>B. You've removed your approval from B. Say lots of people share your
preferences, and they all react to the addition of C in that way, thereby
dropping their approval of B.
That could result in some candidate not in that clone set to now have the
most approvals.
So the addition of a candidate to the clone set caused the winner to no
longer come from the clone set.
So, with the 3-slot modification, MDDA fails ICC. From what Kevin was
saying, MDDA didn't meet ICC to begin with, but neither does it meet it with
the 3 slot limitation.
The 3-slot limitation causes the loss of SFC compliance.
If we use votes-only ICC, MDDA3 might meet it, but so does Plurality, and so
the distinction of meeting ICC is reduced.
So, the 3-slot limitation loses SFC without gaining ICC.
And why not give people the freedom to vote all their preferences?
It seems to me that 67%MDDA fails SFC too. The majority preferring the CW to
Y might be only a 66% majority. Y doesn't get disqualified. Say Y has the
most approvals.
Now, first of all, anything that I say about something immediately after I
find the postings could be incorrect, though what I'm saying here seems
correct.
Also, of course I can't say for sure that the 3-slot modification doesn't
bring soime benefit that's worth losing SFC. So I'm just saying that, at
least based on how things look now, I prefer MDDA & MMPO in their
ranking-unlimited versions, because of SFC.
For me, the choice is between MMPO's LNH, and MDDA's SDSC. (since they both
meet SFC).
I prefer SDSC to LNH because failing LNH won't cause the
preference-concealing voting that I don't like, while passing SDSC can help
reduce those problems.
Also, when no one can be disqualified, or when the most winnable
unacceptable candidate doesn't have the other unacceptables as his biggest
defeats (so you don't need to concern yourself with trying to give him a
majority defeat by bottom-ranking him or using power truncation), you don't
need power-truncation against him, because merely by not approving him, you
give him the worst Approval score you can give him.
So often, maybe usually, the voter is less likely to need power truncation,
in MDDA as opposed to MMPO. But, because sometimes it helps, even in MDDA,
I'd prefer to have power truncation as an option.
As I said, this is a hasty posting, written immediately after finding the
discussion about MDDA3,
67%MMDA, and 67%MDDA3.
Mike Ossipoff
_________________________________________________________________
Dont just search. Find. Check out the new MSN Search!
http://search.msn.click-url.com/go/onm00200636ave/direct/01/
----
Election-methods mailing list - see http://electorama.com/em for list info
