[EM] James, 4/11/05, 2227

[MIKE OSSIPOFF]

James--

You're going to drag your feet all the way, aren't you.

I'd said:

>So no, the problem that could result from offensive order-reversal in wv
>is not a new problem, but is only the old problem, much reduced.


You replied:

	Going from non-burying vulnerable methods (like plurality, runoffs or
IRV) to burying-vulnerable methods (like Condorcet, Borda, Bucklin), the
burying strategy is a new problem in the following sense: In some
scenarios, IRV will pick the Condorcet winner, and strategic incursion
will not be possible

I reply:

James, I'd  just finished explaining to you that the fact that IRV isn't 
vulnerable
to offensive strategy is meaningless, because IRV can have the same kind of 
undesirable results that
offensive strategy in wv would cause. IRV can fail to elect a CW, and IRV 
can
violate majority wishes. Those are the things that offensive strategy can do 
in
wv, but IRV does them without any offensive strategy being used.

We're going around in circles, because you're repeating things that I've 
just
finished answering.

You continue:

... whereas strategic incursion (resulting in the
election of someone other than the sincere CW) will be possible in
Condorcet-efficient methods (e.g. WV).

I reply:

And IRV will elect someone other than the CW without "strategic incursion".

You continue:


	In IRV, given sincere preferences A>B>C>D>E..., one cannot increase the
chances of A winning by manipulating preferences after A, one cannot
increase the chances of B winning by manipulating preferences after B, and
so on. In some situations, this prevents strategic possibilities from
opening up.

I reply:

As I just finished telling you, even though Plurality and IRV aren't 
"vulnerable to
manipulation",  IRV and Plurality will often, without any manipulation, 
produce the undesirable
results that manipulation can bring in wv.

I refer to the results of failing to elect the CW, and violatiing majority
wishes.

You continue:

	In IRV, if candidate X is the sincere Condorcet winner, and more than 1/3
of the voters rank X in first place, then I think that no strategic
manipulation will be possible

I reply:

How many times have I agreed that IRV is not vulnerable to strategic
manipulation? What just doesn't seem to be reaching you is the fact that,
without any strategic manipulation, IRV will give the same kind of 
undesirable results
that strategic manipulation can cause in wv. The undesirable results that I
refer to are failure to elect the CW, and violation of majorilty wishes.

There are 2 differences:

1. IRV will do that without strategic manipulation, but, in wv, those 
undesirable
results are the result of someone trying risky offensive strategy, and 
succeeding with it.

2. Those adverse results can be prevented in wv  by mere equal-voting, as is
also the case with Approval. And, in wv, the "manipulation" that can cause 
that
problem is easily deterred by not voting for the candidate of the expected
manipulators. In contrast, in IRV, those adverse results can be prevented 
only
by favorite-burial.

You stated a condition under which IRV won't have its problem. Of course no
method has its problem all the time. No one ever denied that sometimes IRV
voters will get lucky and not regret sincere voting. I was talking about 
what
methods can do when they go wrong. I though that was what you were talking
about.

You continue:

you open the door for new problems, new liabilities.
The point isn't just that Condorcet doesn't always elect the Condorcet
winner in strategic situations (obviously IRV doesn't either, you don't
need to tell me that), but that vulnerability in Condorcet occurs in some
situations where IRV is invulnerable.

I reply:

As I'd just finished saying in the message that you were "replying" to, 
different methods can give different results with the same set of ballots. 
That's why they're called different methods. Different methods, therefore, 
can have their problems in different examples. If you feel that it means 
something that IRV's failure to elect a CW, or IRV's violations of majority 
wishes happen in different examples from the ones in which that could happen 
in wv, then I answer that that's just something that you should expect, 
because they're not the same method.

This isn't complicated:

I'll use the term "undesirable result" to mean a failure to elect a CW, or a 
violation of majority wishes.

If wv has an undesirable result, that isn't a new thing, because Plurality 
and IRV will often have undesirable results.

You can call it a different problem when a different method has the same 
sort of undesirable results. A "new problem" because it's a new method, and 
the undesirable method happens by a differnent mechanism, due to the 
different methods having different balloting &/or count rules.

So, in that sense, it's a different problem. Let's not use the word 
"problem". Let's instead discuss undesirable results, as I've defined that 
term. Failures to elect a CW, or violations of majority wishes, another term 
that I've defined here.

But, even though you call it a different problem because the undesirable 
result happens with a different method, by a different mechanism due to the 
two methods' different rules, there's something that isn't new or different: 
The undesirable result is the same sort of undesirable result. The 
undesirable result is stil a failure to elect a CW, or a violation of 
majority wishes.

I"m talking about those failures to elect a CW, or violations of majority 
wishes. When that happens, it doesn't matter which method it happened in. 
The CW still lost, and majority wishes were still violated.

But are there still differences in how that happens in Plurality, IRV, and 
wv? There sure are:

1. With Pluralityl and IRV those undesirable results happen automatically. 
With wv they could happen if someone attempts and succeeds with the risky 
offensive strategy of offensive order-reversal.

2. With Plurality and IRV those undesirable results can often be prevented 
only by favorite-burial. With wv, they can be prevented by mere 
equal-ranking. And the offensive strategy that could cause those undesirable 
results is easily deterred by truncation.

So, something is the same with both methods: The possibility of failure to 
elect the CW or a violation of majority wishes. And something is different: 
Those two differences that I numbered and listed above.

When wv adds a possible offensive strategy that Plurality and IRV don't 
have, then, wv is not adding a new violation. It is merely making an old 
violation more difficult to achieve, and easier to prevent.

Because the significant thing, the failure to elect the CW or the violation 
of majority wishes, is the same with all methods when it happens, you're 
creating an artificial distinction when you compare methods by their 
vulnerability to what you call "burying", one particular mechanism for such 
a result. If that were a meaningful distinction, then it would be more 
difficult to compare methods, because you could bring out a whole set of 
possible offensive strategies. But the violations that can happen, the 
undesirable results that can happen, are exactly the same in all of those 
methods. If the CW has lost, or majority wishes violated, that's true 
regardless of what method, by what mechanism, gave that result.

A more meaningful way to compare the methods, then, is according to what it 
takes for those undesirable results to happen, and what it takes to prevent 
them.

Those are the things that I've been discussing. For instance, the majority 
defensive strategy criteria are about what it takes to avoid those 
violations.

FBC doesn't meantion CW or majority, but if no one ever needs 
favorite-burial to get their best outcome, then no one ever needs 
favorite-burial to improve their outcome by protecting the win of a CW or 
avoiding a violation of majority wishes. In that sense, FBC too says 
something about defensive strategy need.

Mike Ossipoff

_________________________________________________________________
Don’t just search. Find. Check out the new MSN Search! 
http://search.msn.click-url.com/go/onm00200636ave/direct/01/