Truncation Resistance #2 criterion (was Re: First Choices tiebre

[Mike Ossipoff]

Hugh Tobin writes:

[Mike writes: This is a re-writing & re-sending of a reply
that I sent yesterday, via the "reply" option. But afterwards
I noticed that only Hugh's address was in the "reply-to" line,
and that my letter therefore didn't post. Yes, I'm still on
the list, at least long enough to answer this]

> 

Regarding the following sentence, "censorship" is a strong
word, which, it seems to me, is being mis-applied. Don, Demorep
& Hugh can say whatever they want, as long as they don't
offer money to a politician in return for a vote for or against
a specific bill, etc. But any group can collectiely define its
purpose, and that isn't censorship. 

> Steve's proposed consensus censorship raises the questions: (1) Why is
> truncation resistance more important that resistance to order-reversal?

Because a) Truncation has taken place in every rank balloting
election of which I'm aware, on a large scale when there are
numerous alternatives, as was the case in our own EM sw methods
vote, while order-reversal is relatively rare (only 1 person
order-reversed in our elecion), and has been agreed to be
unlikely in public political elections on a significant scale;
& b) If you know of a pairwise method that's, in principle,
completely invulnerable to order-reversal, do share that 
information with us. Order-reversal is, so far as anyone
knows, impossible to (Âïøeliminate in principle as a possibilitly
in Condorcet Criterion methods. Therefore its elimination
would seem an unreasonable requirement for a criterion, no?
:-)  However, since offensive truncation strategy can be
made impossible to succeed, that does seem a reasonable
thing to ask.


> (I note that Mike defined both as "drastic" when used as defensive
> strategy; in either case the strategic voter risks giving victory to his

No. I never defined truncation as drastic. I defined drastic
defensive strategy as defensive strategy that ranks a less-liked
alternative equal to or over a more-liked one. Not ranking
someone at all doesn't come under that definition. You could
argue that by not ranking someone, you're effectively ranking
other, less-liked, unranked alternatives equal to him, but
I feel that that's distorting the meaning of "rank". If you
don't rank someone you don't rank him equal to anyone--because
you don't rank him. But by the intent, as well as the
literal meaning says the same thing: Not ranking someone
you like less than your "lesser-evil" is very different
from ranking that lesser-evil equal to or over your favorite.
It's obvious which of those is drastic & self-betraying.

>In either case the strategic voter risks giving victory to  his

> least favorite, as Steve has explained). (2) Why it it critical to

It isn't clear whether you're talking about offensive or defensive
strategy. Yes, in Condorcet(EM), an offensive strategist risks
giving the election to his last choice, if non-drastic countermeasures
are used by his intended victims, if his intended victims don't
co-operate by voting his favorite over his last choice.

But if you're talking about defensive strategy, let me emphasize
that no defensive strategy is needed against truncation in
the method recommended by EM.


> prevent a putatively organized and devious group of plurality voters
> from achieving through truncation what they could also achieve by
> insincerely dividing their second choices -- or what half of them could
> achieve by reversal, if one assumes the true second choice of all of
> them is the middle candidate?

Partly because, as I said, truncation will be, & has been, 
common, while order-reversal is rare. And if, as you say,
that orchestrated split order-reversal would have the same
effect as truncation, it isn't clear why it would be
done instead, or why you believe it needs a separate
deterrence.

> 
> I submit that a system which reaches a different result from Steve's
> second example if the "A" voters are divided into 23AB and 23AC creates
> a clear incentive to insincere voting.

Submitting it & demonstrating it aren't quite the same thing.
What you suggest would be order-reversal by half of the A voters
(since you call it insincere). There's reaally no reason to
do order reversal, & to intentionally only do it halfway.

Look, Hugh, we've been all over this. You didn't answer my
replies to your claims when we discussed it before, and now,
for at least the 2nd time, you're repeating your statements
again. Disorderly discussion. We covered that topic very thoroughly.
I understand that EM is archived, and so I suggest you refer to
my comments when we were discussing that topic before. This
isn't the first time I''ve had to suggest that to you.

> 
> Given that incentives to strategic voting exist in all systems, it seems
> the burden is on Steve to show why the incentive that exists if A wins
> in example 2 is presumptively worse than those that may exist in a
> universe of unspecified systems.

For one thing, you're comparing two things that don't match:
incentives with systems. You want Steve to show that an
incentive is worse than a system? (or a universe of systems?)
Say what you mean.

Do you mean that incentives in the examples must be shown to
be worse than incentives in "a universe of unspecified systems"?
A tall order :-)

Anyway, as I understood the example, it didn't speak of 
incentives, but only of results. If you want to make
a proposal-restriction that specifically refers to
incentives, then _you_ do so.

If you have a better suggestion for a criterion or a requirement
for proposals, then do trot it out.


> 
> -- Hugh Tobin
> 
> Steve Eppley wrote:
> > 
> > Demorep wrote:
> > >After rethinking the problem of Condorcet circular ties, I note
> > >that such ties occur because any tiebreaker method using additional
> > >rankings beyond the first choice ranking has the potential for
> > >strategic voting
> > -snip-
> > 
> > The Gibbard-Satterthwaite theorem points out that *any* method has
> > the potential for strategic voting.  Demorep's proposed new method
> > is even worse than Condorcet and Smith//Condorcet.
> > 
> > I suggest that when people think about a new method they should
> > first test them on the following two examples:
> > 
> >      46:ABC                          46:A
> >      10:BAC                          10:BAC
> >      10:BCA                          10:BCA
> >      34:CBA                          34:CBA
> >      ------                          ------
> >      MUST ELECT B                    MUST NOT ELECT A
> > 
> > In the example on the left, it should elect B (the compromise
> > centrist who would beat any other in a head to head matchup).  In
> > the example on the right, it should not elect A, else the supporters
> > of A have a clear incentive to strategize.  If truncation would
> > elect A, the method has a serious problem.
> > 
> > We may as well call this two-example test a criterion: the
> > "Truncation Resistance #2" criterion, perhaps.  Do we have a
> > consensus that any method which fails TR-2 should not be posted
> > here in EM, unless it's accompanied by a solid explanation why it's
> > better than Condorcet or Smith-Condorcet on some other criterion we
> > should respect?  Such a consensus would save us all a lot of time.
> > 
> > If I understand it right, Demorep's new "First Choices tiebreaker"
> > method fails TR-2.  Though it passes the left example (elects B),
> > it fails the example on the right: the circular tie is resolved by
> > eliminating B (which has fewest first choices), then A beats C.
> > 
> > ---Steve     (Steve Eppley    seppley at alumni.caltech.edu)
> 
> .-
> 


--