[EM] Blake reply

MIKE OSSIPOFF nkklrp at hotmail.com
Mon Apr 30 23:40:56 PDT 2001


I'd said:

>I notice that you
>haven't posted any such examples. It couldn't be, could it, that
that's
>because you don't have any examples in which those wv strategy
"problems"
>cause serious consequences?


Blake said:

When you talk about "serious consequences", you mean from a global
perspective.  In other words, you are referring to situations where
you don't approve of the outcome.

I reply:

That's nonsense. The serious consequences that I demonstrated would
be disliked by a majority of the voters, not just by me.

Blake continues:

When I talk about voters being punished, I mean from the perspective
of the individual voter.  That's the perspective that actually
motivates how a voter votes.

I reply:

Margins can make a majority-size group of voters regret that they
didn't vote for the sincere CW, voted over their candidate by
a majority. That result isn't dependable enough to deter the truncation
as an offensive strategy, but it will of course happen sometimes.
And so Margins, as I pointed out in my previous letter, is the
method that punishes truncation. Of course with any pairwise count
method, the A voters need to rank B if C beats A. Under those conditions, 
they don't want to let C beat B also. But if A beats C,
Margins will still sometimes make the A voters regret their
truncation of sincere CW B, even when A beats C.

I thought it was obvious that my concerns are from the perspective
of the individual voter. I've always made it clear that my concerns
are of that nature.

In my examples 1, 2, & 3, the individual voter who's a member of the
B>A majority will regret that he didn't use drastic defensive strategy
to defeat A.

I'd said:

>Thank you, but would you show some examples showing how wv produces
>seriously bad results, results as bad as the Margins results that
>wv gets rid of?

Blake obliges, and says:


45 A B C
20 B A C
35 C B A

If the A 1st voters randomly rank after A, and split evenly, then A
will win, which is the kind of bad result that concerns you.  So,
that's one reason I say wv doesn't do any good.  Voters can avoid the
penalty by randomly ranking.


I reply:

Nonsense. If your above rankings are sincere preferences, then those
22.5 A voters are order-reversing. But sure, it's easy to make
examples, like my example 3, in which the A voters are indifferent
between B & C, and they nevertheless vote C over B. It's, as I said,
a weaker relative of order-reversal.

Please note that A wins if Blake's example is counted by Margins too.

The difference? That falsification can win in Margins even if the
B voters don't vote for A. Blake had to make the B voters rank A
in 2nd place, helping their victimizers, in order for the falsification
to be able to steal the election in wv. As I said, with wv,
falsification, including order-reversal, can't steal the election
from a candidate whom a majority vote over the falsifiers' candidate
unless the victims of the strategy trust & help their victimizers.

Blake continues:

If you manage to convince people that truncation in this example is a
major strategy problem for margins, reasonable people will conclude
that random-ranking is also a major strategy problem for wv.

I reply:

In your example, it's order-reversal. But yes, falsification among
equally-preferred candidates can steal the election as I've described
with both methods. But with Margins, it can do so even if the B voters
don't vote for A.

And, as I said, truncation, which is always widespread in rank ballotings, 
will steal the election as I described in Margins, but
can't possibly do so in wv. That's why we say that wv is truncation
resistant.

Blake, does that example represent your best effort to show a problem
for wv?

I'd said:

>What isn't considering it. I'm considering it. That's one reason why
>there can be truncation. Other reasons are strategic, and lazy, and
>principled--I'd refuse to rank anyone unacceptable, just as I'd
refuse
>to rank an unacceptable voting system in a voing system poll.

Blake said:

Actually, I was referring to your examples where people truncate
against their best interests, either out of spite, or because it makes
your example work.

As for these various reasons for truncation, I think you may be going
beyond a reasonable concept of a sincere CW.

I reply:

Sincere CW:

A candidate who, when compared separately to each one of the others,
is preferred to that candidate by more voters than vice-versa.

[end of definition]

That's my definition of a sincere CW. How would you define the term?

Blake continues:

For example, I am
concerned that factions might use strategy to affect the election
result, including by defeating the sincere CW.  But you seem to want
to preserve people's preferences even if they can't be bothered to
express them, or if they don't want to express them, independent of
strategy concerns.

If someone can't be bothered to rank B over C, I don't consider the
information that they actually prefer B to C to be very important.  I
don't think an opinion is likely to be well thought out, if the voter
can't even be bothered to mark it on the ballot.

I reply:

Maybe there are 20 or 30 candidates, and the voter has other things
to do.

Actually, whether those voters who don't rank B really are indifferent
between B & C, or prefer B to C but are insincerely truncating,
doesn't change anything. Either way, if A majority vote B over A,
and no majority vote anyone over B, then A can't win. Under those
conditions the B>A defeat can't be the weakest defeat in a cycle,
and that means that in SSD, Cloneproof SSD, and RP(wv), A can't
possibly win.

I wanted to word SFC in terms of a sincere CW, and we wanted to word
GSFC in terms of the sincere Smith set, in order to write them about
familiar situations that guarantee the required conditions. Also,
it gave brevity.

Also, the approach of sincere preferences & sincere voting
avoids the farsical embarrassment of Plurality meeting the criterion,
or the shabbiness of having to say that the criterion doesn't apply
to Plurality or Approval, as Blake has to say about his Condorcet Criterion.


Blake continues [referring to truncation]:

It's an even better strategy in both, not just margins.


I reply:

Yes, because truncation can't possibly succeed in wv under the
plausible conditions I described, that makes order-reversal
the only offensive strategy. In Margins, order-reversal may be
more effective than truncation too. Note how little is required to
successfully order-reverse in Margins, as shown in my order-reversal
example. Even if B is nearly as big as A, the A voters can merely
pick a candidate less than half as big as A, and which A beats, and
rank him over B, and then, even if B voters don't rank A, A wins.

So yes, Margins is even more vulnerable to order-reversal than to
truncation. But, as I said, truncation happens all the time--though
of course order-reversal will start happening all the time if
Margins is enacted.

Blake continues:

The reason you truncate, is that you have a victory, like X>Y, and you
want to increase it.  At some point, the victory gets to  be enough to
defeat Y.

I reply:

No, maybe without the truncation Y beats X, and after the truncation
X beats Y and your favorite candidate wins, in Margins.

Blake continues:

Then, either one of the truncator's preferred candidate's
wins, or the strategy back-fires.  The only reason you would truncate
is if you expect a better result when Y is sufficiently defeated.  If
that is the case, it makes sense to do all you can to sufficiently
defeat Y.

I don't deny that order-reversal is more ridiculously easy in Margins,
as demonstrated in example 2.

But truncation happens a lot already, and it too will sometimes
steal an election in Margins.

Blake continues:

So, we're actually better off if people believe you when you talk
about how great a strategy truncation is.  This prevents them from
using more sophisticated strategies.

I reply:

I'm sorry to disappoint you, but if Margins were used in public
political elections, I'd be making people aware of Margins'
ridiculously easy order-reversal success conditions, when those
conditions exist and when that order-reversal would improve the
outcome for me.

My concern about truncation is due to its common incidence in all
rank ballotings, and Margins' vulnerability to it.

Mike Ossipoff



_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com



More information about the Election-Methods mailing list