[EM] Dave: We agree about the first step. Jameson: RP vs Beatpath

[Michael Ossipoff]

Jameson:

I don't know the criterion-compliance difference between RP and Beatepath.
Some years ago there was acrimonious
debate here between their advocaters, and it seems to me that they did
speak of different criterion compliances.

So far as I know, though, the main differences are:

1. Beatpath is more easily implemented, in regards to program-writing and
computation-time. (not that either would
need significant computation-time).

2. RP is much more briefly defined and explained.

Beatpath has a good explanation and motivation, when defined and explained
in terms of
Schwartz-Sequential-Dropping (SSD). Beatpath and SSD are equivalent when
there are no
pairwise ties. Pairwise ties are rare with large electorates. A versioni of
SSD, called CSSD, is
exactly equivalent to Beatpath, duplicating its results even in small
electorates, with pairwise ties.

For public elections, of course the simpler SSD would work fine.

I'd define SSD here, but what would be the point? These methods fail FBC,
and aren't defection-resistant.

I know I've said these things before, but I want to say them again, because
Adrian wasn't on this list
when I said them before:

FBC:

When the presidential election included Ralph Nader, John Kerry, and George
W. Bush, all the Democrats would say,
"Make your protest vote for Nader some other time, because this time it's
crucial that the Republican not win. To avoid
disaster, vote for Kerry, even if you prefer Nader."

What if we had Condorcet's method? Someone surely would mention this. You
couldn't keep it secret. They'd say:

"If you vote for Nader and Kerry in 1st place, Bush might win, where Kerry
would have won if you'd ranked him _alone
in 1st place. To avoid disaster, we must do _everything_ we can to help
Kerry beat Bush. To _fully_ help Kerry against
Bush, we've got to all rank Kerry alone in 1st place."

That's true of Condorcet, but it isn't true of ER-Bucklin or ICT. No one
could make the above statementy about dsER-Bucklin
or ICT.

But, even with ICT,  would the voters really believe that it's safe to
top-rank both Nader and Kerry? Maybe not, because
all rank methods are complicated. With Approval it's quite plainly obvious
that you can fully help Kerry against Bush by
approving Kerry, giving him 1 point, and giving Bush  0 pointgs--and you
can do that while also giving Nader 1. Giving a
point to Nader, obviously, in no way affects your helping Kerry against
Bush.

Defection-Resistance:

Here's an example that I call the "Approval bad-example". It's a (solvable)
problem for Approval, but also, just as much
for Condorcet. It's Approval's only problem that is worthy of the name. It
has a number of good solutions in Approval. No
doubt those solutions
 will also work in Condorcet. But the point is that Condorcet fully shares
that problem of Approval.
ICT doesn't have that problem.

Here is the example:

Sincere rankings:

27: A>B (meaning that they prefer A to B to C)
24: B>A
49: C (meaning that they're indifferent between everyone but C)

Now, in Approval, if the A voters and B voters approve eachother's
candidates, then the winner will be
one of {A,B}. It will be A.

But what if the a voters approve B, and the B voters don't approve A? B
will win, having taken advantage
of the A voters' co-opreration.

The A voters could try to protect themselves against that, by refusing to
approve B, but then, if neither
approves the other, C wins.

Someone has to co-operate, or C wins. But the co-operators are giving it
away to the defectors.

The message is, "You help, you lose."

There are 5 ways that that problem can be well dealt-with and solved in
Approval, and probably
in Condorcet too. I'll find and re-post the posting in which I listed them.

Here's the defection scenario:

27: A>B (They rank A in 1st place, and B in 2nd place)
24: B (They rank only B)
49: C

Condorcetists are in denial about this problem. It's as fully present in
Condorcet as in Approval.

It isn't present in ICT at first strategy level (and only minimally at 2nd
level). ICT is defection-resistant.
So much so that it maybe can be said to be defection proof. I'll define ICT
in a posting today.

Jameson:

We don't have any disagreement about the 1st step: Approval.

When Approval has been in use for a while, people will be more familiar
with voting systems, and very likely
we'll have a more open media, with better honesty and better media-acess.
Under those conditions, the
rank-balloting contraptions will have an enactment chance. Maybe eventually
people won't be inclined
to bury their favorite in Condorcet. But even if so, Condorcet will still
fail to be defection-resistant (see above).

So, some time after the enactment of Approval, we can argue ICT vs
Condorcet. I'll be arguing that we should
either keep Approval, or enact ICT.

But that issue isn't a problem now. If you want to work for Condorcet, the
way to do that now is by helping
to enact Approval.

Mike Ossipoff


Mike Ossipoff


.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20120423/452de9c7/attachment-0003.htm>