[EM] MDDA vs UncAAO, ASM & DMC

[Michael Ossipoff]

Forest--

I’d said:

UncAAO, ASM, and DMC ignore defeat-magnitude and magnitude of pair-wise 
opposition. They look at pair-wise vote totals in order to find out who 
beats whom, but then they throw pair-wise vote total information away. 
Throwing that information away has regrettable consequences.

You replied:
:
Making use of pairwise defeat magnitudes is a two edged sword. Yes, it gives 
information, but it also gives incentive for voters to go out of their way 
to add more to the defeat strengths that they feel a strategic need for even 
when the defeat itself is already assured.

Substituting Approval measures of defeat strength allows for strategic 
control without requiring order reversal or collapse.

I reply:

But strategy is then more often needed than with MDDA, MAMPO or wv 
Condorcet.

I have nothing against Approval, but relying on it to measure 
defeat-strength brings Approval’s strategy need to rank methods. Rank 
methods could offer more than that.

Offensive order-reversal works in DMC too. If there’s less incentive for it, 
it’s because truncation works just as well in DMC.  DMC is vulnerable to 
truncation in a sense that MDDA, MAMPO and wv Condorcet are not, as examples 
below will demonstrate.

You spoke of incentive for order-reversal. I’m more interested in strategic 
_need_ for order-reversal. That need is at its worst when a voter needs to 
bury his favorite in order to keep a greater-evil from winning, to protect 
majority rule, or to protect a CW. That need will never exist with MDDA or 
MAMPO. But it will with DMC. There are numerous possible examples.

   51: AC (offensive order-reversal)
100:   BA
50: C/B  (“/” denotes approval cutoff)

The A voters have made a cycle, and, B, being least approved, is eliminated, 
and A wins.

But if one C voter voted B over C, then B would win its pair-wise comparison 
with C, and so B would win the count.

All of the truncation examples that follow are also FBC failure examples, 
since one C voter could reverse the C>B defeat by voting B over C.

As I said, DMC has a problem with truncation, where MDDA, MAMPO, and wv 
Condorcet do not:

102: A
100: B
101: C/B

The A voters’ truncation causes a cycle. B again has the least approval and 
is eliminated, electing A.

As before, one C voter could save the CW, enforce majority rule, and defeat 
a greater-evil by voting B over his/her favorite.

I emphasize that the need for favorite burial is nonexistent in MDDA and 
MAMPO.

You can object that the above problem happens because the C voters didn’t 
approval B. Sure, but that’s the point: Strategy use is needed to enforce 
majority rule, protect the CW, and prevent the election of the greater-evil. 
That isn’t so with MDDA, MAMPO or wv Condorcet. With those methods the B 
voters and C voters need do nothing other than rank sincerely, in order to 
keep A from winning. For that reason, MDDA, MAMPO and wv Condorcet fulfill 
the full promise of rank balloting.

    2: A
100: BA
101: C/B

This time, with the B voters helping A, even two A voters can steal the 
election from the CW by truncation. Yes, the B voters and C voters could use 
better approval strategy, thereby saving the CW, enforcing majority rule, 
and preventing the election of a greater-evil.. But, as I said, that’s the 
whole point: They need to use the right approval strategy. In MDDA, MAMPO 
and wv Condorcet, they don’t need any strategy at all. They need only rank 
sincerely, and it’s guaranteed by SFC compliance that A can’t win.

Mike Ossipoff