[EM] Criteria

[Kevin Venzke]

Mike,

Your definition of the scheme is not bad.

> It’s far from obvious what a FARCS 
> criterion has to do with actual voting and strategic concerns.

It's not clear to me how one system of criteria could have more
or less to do with "actual voting and strategic concerns" than
another.

> Such a 
> criterion speaks of rankings, something that is meaningless for 
> a method that doesn’t use rankings.

I wouldn't call it "meaningless" since you can still apply the
criteria. Define a criterion for instance that says "the candidate
with the highest average voted rating must win." It can still be
applied to a method that doesn't use ratings; it's just that this
method is not likely to satisfy it.

> In what sense do you mean that a voter in Plurality 
> intended a ranking?? If it isn’t clear to you what that means, it 
> isn’t just 
> you--it really is not clear. FARCS advocates on EM haven’t succeeded 
> in clearly answering that.

I don't believe it needs to be answered. That is a difference from
working with sincere preferences. I don't have to make any assumptions
about why the voter intends to vote some ranking, whether or not
the method even supports this intended vote.

In the same way, writers like Woodall don't speculate about whether
voted ballots bear any relation to some sincere preferences behind them.
It just doesn't matter. If there are sincere preferences behind the
votes, then the criteria will make useful guarantees. If not, then they
won't. There's no need to ask whether a rational voter would ever have
reason to provide some vote.

> But how about SFC? SFC involves the CW. The CW is defined in terms of 
> preference. You could try substituting  “beats-all candidate” for CW.
Would 
> that work? With any pair-wise count method, the beats-all candidate wins.

> That means that candidate over whom a majority rank the beats-all
candidate 
> must lose. So every pair-wise count method would meet votes-only SFC, if
we 
> use “beats-all candidate” for CW.
> And, if we don’t use “beats-all candidate” for CW, then how do we
interpret 
> “CW”, if we aren’t allowed to mention preferences?

You have to consider what the implications of SFC are when you only have
the voted ballots and have to speculate about what the sincere preferences
behind them were. A votes-only version of SFC (defined by Markus IIRC) is
that if A has a majority over B and there is no majority over A, then B
doesn't win.

As far as A having a majority over B, SFC already says as much. As far as
there being no majority over A, SFC basically implies this by saying that
A is the sincere CW and no one is using order reversal.

> So, in addition to FARCS’ other problems, can FARCS criteria even be 
> equivalent to preference criteria?

Not necessarily, no. Votes-only interpretations of FBC and SDSC are not 
"equivalent" to FBC and SDSC.

> Another thing: When  your criteria are written in terms of rankings, and 
> Plurality fails, you’re not in a good position to say that the criterion 
> isn’t merely acting as a “rules criterion”. Maybe Plurality fails because

> it doesn’t have the kind of balloting on which your criteria are based.

But methods don't simply fail criteria because they don't use the balloting
system of the criterion. For example, a method that allows truncation but
not equal ranking could satisfy MD but not likely FBC. This is a result of
how the scheme works; it's not the same as saying "if you don't allow equal
ranking then you fail FBC."

> I’ve been told that FARCS is the standard 
> academic approach to criteria involving preference, as their way of
trying 
> to avoid preference..

I don't think this is true. E.g. Woodall doesn't use a distinction between
an intended ballot and the ballot as allowed to be expressed. He also
doesn't apply criteria to non-rank methods, although he uses ranked ballot
interpretations of FPP and Approval.

> Do you tell your imaginary rankings story with each criterion-definition 
> (if 
> so I haven’t noticed that being done), or do you just expect it to be 
> understood that criteria mean something different from what they say?

I expect a lot of it to be understood, yes. It's not usually a problem.

> You continue:
>
> And I showed e.g. why FPP fails minimal defense.
> 
> I reply:
> 
> Using FARCS? Not when I asked for an application of FARCS..

Yes. I'm not sure what you mean when you say "when I asked for an 
application of FARCS" since in April '05 this isn't what you were asking
me for.

> Must we still say that no FARCS advocate has ever defined and applied
FARCS 
> on EM?

Your definition is very similar to what I gave in my Feb post.

> Maybe Plurality fails FARCS Condorcet Critrerion because FARCS 
> is about rankings and Plurality doesn’t have rankings. A rules criterion.

Plurality fails Condorcet *indirectly* because it doesn't allow full
rankings. It isn't the same as saying this is the very reason that it
fails.

> You’re actually calling FARCS’ roundabout fictitious tangle a simpler 
> system??

Simpler to work with, yes.

> First of all, if your criteria system doesn’t apply to Approval and 
> Majority Favorite, how can you call it acceptable?

There is no way to define Approval so that it satisfies MF under my
scheme. It would be nice to be able to do better than assert this, but 
I think it is intuitively obvious that there are only a few ways to
treat Approval.

> I haven’t checked this, but let’s suppose that this, with FARCS, is 
> equivalent to SFC:
>
> If there’s a candidate who doesn’t have a majority pair wise defeat, 
> then no one who has one should win.

Just FYI, this would not be satisfied by e.g. Schulze(wv) or MAMPO.

> So, actually, your FARCS criteria are simple and brief only until we 
> try to apply them to nonrank methods.

I don't have a strong objection to this opinion.

Kevin Venzke


	

	
		
___________________________________________________________________________ 
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 
http://fr.answers.yahoo.com