[EM] Jobst: Strategy-Free Criterion

[MIKE OSSIPOFF]

EM Members--

Sorry to break my retirement promise again, but this time it will just be 
temporary. At almost  the exact same time that I retired from voting 
systems, a Range-Voting advocacy organization was formed, and I couldn't 
resist taking part to some small extent. In for a dime, in for a dollar. For 
me, either I'm completely out of something, or I'm not out of it at all. So, 
I made a short postponement of my retirement.

So I was browsing the EM archives, and I noticed a message posted in March, 
which I'd missed at the time. So I'm going to reply to it now:

Dear Jobst--

I'd said:

>Here´s the actual definition of SFC:
>
>SFC:
>
>If no one falsifies a preference, and if a majority prefer the CW to
>candidate Y, and vote sincerely, then Y shouldn´t win.
>
>[end of SFC definition]
>
>Which part of that don´t you undestand?

You (Jobst) replied:

Well, I at least think I understand it, assuming you speak of the
*sincere* CW (otherwise, that is, if you meant CW according to the cast
prefernces, then each Condorcet method would fulfil SFC trivially by
definition, and that cannot be what you meant).

I reply:;

1. I was writing that to someone else. You weren't then a participant in 
that discussion,and I was not questioning your understanding of SFC.

2. The "sincere" CW is the only kind that I've heard of. Though some people 
have, at times, misused the term by applying it to the BeatsAll candidate, 
the definition of a CW is pretty much agreed-upon by all. Though the 
definition is worded in different ways, some more precise than others, this 
is what it always means:

A Condocet winner is a candidate who, when compared separately to each one 
of the other candidates, is preferred to that candidate by a number of 
voters greater than the number of voters who prefer that candidate to 
him/her.

[end of CW definition]

Maybe the wording could be made clearer by naming the candidates:

A Condorcet winner, is a candidate, C,  who, when compared separately to 
each other candidate, X, is preferred to X by a number of voters greater 
than the number of voters who prefer X to C.

[end of alternative wording of CW definition]

Often an author will just say that the CW is a candidate who'd beat each 
other candidate in separate 2-candidate races. That means the same thing, 
though it isn't really as complete.

Anyway, that definition, worded in these several ways, and maybe a few 
others, is the only accepted definition of a CW. So the "sinceree" CW is the 
only CW.

You continued:

But: Can you tell me just one method which passes that criterion?

I reply:

No. But I can tell you a class of methods that pass that criterion: All of 
the wv Condorcet versions pass SFC.

These include the wv  versions of BeatpathWinner, CSSD, SSD, Ranked Pairs, 
MAM, PC, Smith//PC, and SD.

Additionally, one of the recently-described FBC-complying rank methods 
passes SFC.

You continued:

Approval certainly doesn't.

I reply:

Correct: Approval doesn't pass SFC. But Approval passes FBC, which 
Condorcet-Criterion methods don't pass, and which is absolutely necessary 
for today's public political electorates, for whom there's a "greater-evil" 
whose defeat is so important that people feel a need to abandon their 
favorite, if necessary, in order to beat him.

Approval additionally passes WDSC, which not many methods pass. Approval 
also has a valuable optimization property described at the CRV website.

Range-Voting (also called CR, or the point system) likewise passes FBC and 
WDSC, and shares that optimization property.

You continued:

So, although I find anti-strategy criteria most important and are
completely d'accord with you that their formulation will of course
contain references to the sincere preferences of the voters, I still
don't think that this particular criterion, at least not in the above
version, is particularly useful...

I reply:

You seem to be saying that you don't find SFC useful because Approval 
doesn't pass SFC. But then you must not find Condorcet's Criterion or the 
Smith Criterion useful either.

Or is it that you're assuming that no method meets SFC? But I've said many 
times on EM, and at the websites that have my articles, that the wv 
Condorcet versions all pass SFC.

SFC is a majority-personalized version of Condorcet's Criterion (CC). CC 
makes a guarantee to the entire electorate, that if they vote sincerely the 
CW will win. But maybe we can't guarantee that everyone will vote sincerely.

So SFC makes a guarantee to a majority who prefer the CW to some candidate 
Y, stipulating that they, but not necessarily anyone else, vote sincerely. A 
particular prefernce-group can more reasonably assume that they vote 
sincerely, as opposed to assuming that everyone, including the voters they 
disagree with most, will vote sincerely. SFC is majority-personallized CC. 
That's why I prefer SFC to CC, and considcr SFC more "useful" than CC. 
Because maybe not everyone will vote sincerely.

There are several voter-set-sizes to which a criterion can make its 
guarantee. Obviously the best is for a criterion to make a guarantee 
individually to each and every voter. For instance, FBC guarantees, to each 
voter individually, that s/he has no reason to vote someone over his/her 
favorite.

But I wanted to guarantee more, and the additional guarantees that I wanted 
to make can't be made to each individual. But they could be made to a 
majority. Hence the majority defensive strategy criteria, of which SFC is 
one.

Condorcet's Criterion is an example of a criterion makes its guarantee to 
the largest voter-set, the entire electorate, while stipulating sincere 
voting by all.

The more people whose sincerity has to be stipulated, the less useful the 
criterion is, obviously.

FBC is essential for public election methods. But in organizations and 
committees of honest people who aren't too antagonistic to eachother, where 
there's no really scary "greater-evil", FBC isn't so necessary. The wv 
methods can fail FBC, but not easily. Without a scary greater-evil, we 
needn't worry about wv's FBC failure so much. Under those conditions, wv's 
SFC compliance becomes very valuable. To a lesser, but still meaningful, 
extent, its CC compliance becomes useful too. That's why I've been 
recommending BeatpathWinner/CSSD to organizations and committees for a long 
time.

Of course if an organization wants to use, for its own elections or choices, 
a more publicly winnable method, in order to demonstrate the method or help 
set a use-precedent for it ,then RV would be the best choice.

Mike Ossipoff

_________________________________________________________________
On the road to retirement? Check out MSN Life Events for advice on how to 
get there! http://lifeevents.msn.com/category.aspx?cid=Retirement