[EM] The Repoman strikes again

[LAYTON Craig]

Mike wrote (in part):

>The problem is that you don't seem to know what you mean by "procedure".
>
>As you use the term, are Condorcet's Criterion, Smith's Criterion,
>& Monotonicity about procedure or outcome? They're certainly about
>outcome, but maybe you'd say they're about procedure because
>a procedure's merit can be judged by outcome. In that case you're
>all confused about what you mean by procedure.
>
>So are you saying that we should drop Condorcet's Criterion, Smith's
>Criterion, & Monotonicity, because they're about outcome rather than
>procedure?
>
>If you say they're about procedure, & therefore ok, and if you
>agree with Condorcet's Criterion & Smith's Criterion, then can
>you notice that, in Demorep's example, IRV is procedurally acting wrong
>(because you consider Condorcet's Criterion to be about procedure)?

Yes, of course it is.

>Or is it that you're still bothered by the fact that Demorep used
>Washington for the voter median candidate, and Stalin & Hitler for
>the extremes. 

And yes I'm still bothered by this.

>In the U.S., a someone more palatable than Hitler &
>Stalin, say maybe Washington, would be more likely to be the voter
>median candidate. So Demorep used names that accurately reflect
>who would be an extreme and who would be more likely the voter median
>candidate. 

Soviet-aligned (ie aligned with Stalin) Communist Party candidates in the
30's in Germany decided to withdraw to support Hitler under the theory that
Nazism was an essential precursor to communist revolution under German
conditions.  Another example that votes to not necessarily flow from
unpalitable 'extreme' candidates towards friendly centrist candidates.

>Did he exaggerate the extremes? Sure. But the point is
>entirely valid, about how IRV fails by jumping to extremes, who are
>a danger of being much less desirable.

You pointed out yourself that the nature of the 'multidimensional issue
space' (ie the Zhirinovsky/Zuganov example) means that who the extreme or
median candidate is doesn't necessarily indicate desirability (perhaps
slightly more likely, if people are considered to be rational utility
maximisers, but then again, perhaps less likely if we consider that
opression or prejudice just as often informs peoples decisions).

>>If you say the latter, you're working backwards.  Say there is a sincere
>>Condorcet winner in an election.  Should that candidate win, even if the
>>voters don't vote sincerely and another candidate is the voted Condorcet
>>winner?
>
>You're using "Condorcet winner" in 2 different ways. First you
>use it to mean sincere Condorcet winner, then you use it to mean
>voted Condorcet winner.
>
>The sincere Condorcet winner should win if everyone votes sincerely.
>If you've believe that I, Demorep, or anyone else has said that
>the SCW should win no matter how people vote, would you please
>state the date of that posting, so we can find it in the archives?

You haven't, but why not?  Isn't the ultimate outcome for the SCW to win
every time?  (If electoral theory is so outcome based, as you suggest).  I
haven't suggested it, because it cannot be incorporated into a fair
proceedure.

The concept of electoral theory being proceedurally based is fairly simple;

1) We have no way of telling who the best candidate will be
2) Because of (1), and for various other good reasons, holding an election
seems to be the best way to decide which candidate wins
3) We develop a proceedure for counting votes and selecting a winning
candidate based on the fairest representation of peoples views as expressed
in (2).

How we determine "fairest representation of people's views" is partly
determined, as you point out, by outcome-based arguments.

Point 1 is the disputed point.  Is this correct?

>>Of course not.  We all think that the fairness of the proceedure is
>>more important than the fairness of the outcome, if only because you can
>>guarantee a fair proceedure with much more certainty than a fair outcome.
>
>This again goes back to your confusion about what you include under
>procedure. Presumably you feel that the criteria that have requirements
>about outcome are really about procedure and not about outcome.
>
>Here's a procedure criterion that we've heard on this list:
>
>The candidate who first ends up with a majority when lowest votegetters are

>eliminated and their ballots transferred according to the rankings
>should win.
>
>Is that a fair procedure or an unfair procedure?
>
>The use of that procedure violates a number of important outcome
>criteria, some of them widely accepted. I'd rather point that out
>than debate the fairness of that procedure.